Professional Documents
Culture Documents
2018 Marketing Science Back Matterfinal
2018 Marketing Science Back Matterfinal
TECHNICAL SESSIONS
n TA01
Room 32, Alter Hall
This Technical Session listing, which provides the Consumer Search
most detailed information. The listing is presented General Session
chronologically by day/time, showing each session and Chair: Rajdeep Grewal, University of North Carolina, Chapel Hill, NC,
the papers/abstracts/authors within each session. 27599, United States, [email protected]
1 - A Model of Online Product Search with Endogenous
Review Sampling
Marton Varga, INSEAD Business School, Boulevard de Constance,
The Session Codes Fontainebleau, 77305, France, [email protected],
Paulo Albuquerque
The authors propose a structural model of online search with endogenous
TA01
product page and review sampling, subject to search costs. Product’s utility to
Room number. Room locations are consumers comes from three sources: (i) product characteristics, displayed in the
also indicated in the listing for each category page and thus known to the consumer before product search; (ii)
session. additional product characteristics revealed upon sampling the product page; (iii)
an experiential-related shock, representing personal satisfaction with the product
upon purchase. While searching, consumers update their beliefs about the
distribution of the third term by reading product reviews and thereby mitigate
Time Block. Matches the time related uncertainty, at a cost. The authors show that there exists an optimal
strategy for product and review sampling and apply the model to study consumer
The day of blocks shown in the Program search and purchase behavior at an online retailer.
Schedule.
2 - Inferring Automobile Market Structure from Search Data
Pradeep Chintagunta, University of Chicago, Chicago, IL,
United States, [email protected],
Time Blocks Jae Hyen Chung, Sanjog Mishra
In this paper, the authors present a new approach that allows inferring the
Thursday market structure from aggregate market share data and individual search data.
Using a new likelihood-based estimator that leverages the joint likelihood of the
A — 8:30am - 10:00am search and purchase decisions, authors obtain market structure while
incorporating consideration sets into the analysis. Market structure is based on a
B — 10:30pm - 12:00pm factor-analytic approach that decomposes the distribution of preferences and that
is embedded in the estimation of the model parameters. The authors apply the
C — 1:30pm - 3:00pm method to a simulated data set to demonstrate that it provides consistent
estimates and apply it to SUV market share data and individual online vehicle
D — 3:30pm - 5:00pm search sequence data to demonstrate its application in an empirical context.
Friday Comparing the results from a multinomial probit model that analyzes the market
structure only with market share data, the authors show the importance of taking
A — 8:30am - 10:00am consideration sets into account when studying the nature of competitive
interactions in the marketplace.
B — 10:30am - 12:00pm
3 - Cost of Webrooming: A Case of Automobiles Market
C — 1:30pm - 3:00pm Guneet Kaur Nagpal, UNC-Chapel Hill, Kenan Flagler Business
School, Chapel Hill, NC, 27514, United States,
D — 3:30pm - 5:00pm [email protected], Rajdeep Grewal
Saturday Webrooming refers to the shopping strategy wherein shoppers first search the
products online and then visit stores to purchase them. A typical shopper in the
A — 8:30am - 10:00am automobiles category is a webroomer, who spends on average 11.5 hours online
B — 10:30am - 12:00pm before visting 1-2 dealerships in the proximity. The authors seek to estimate the
cost of webrooming by first providing an evidence that there exist two mutually
C — 1:30pm - 3:00pm exclusive sequential stages of consumer search in the automobiles category, i.e.
online search followed by travel to automobile dealerships; and then estimating
D — 3:30pm - 5:00pm the consumer search cost in these two stages. For the first stage, which is the
brand-choice stage, the authors estimate the automobile demand in presence of
price uncertainty about the in-market brands, and the modeling approach
rationalizes the observed consideration sets as an outcome of consumers’ online
search effort. For the second stage, which is the dealership-choice stage,the
authors rationalize the price dispersion with the variation in consumer travel
costs. Authors apply the models to the purchase behavior of 70,000 unique
autoshoppers who are webroomers.
1
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 2
n TA02
process production, and whether and when a manager should choose to
automate the production process. Comparing the costs of an automated
production system to one that is not, we show that while automation provides
Room 33, Alter Hall significant cost-cutting benefits, it may still result in an overall costlier system
Digital Economy I: Automation, AI, and because of the increase in the cost of incentivizing the remaining employees. We
Online Technology show that as some tasks in production can be automated and automation
becomes gradually more cost-efficient, human teams are less likely to be
General Session employed under cooperative contracts which reward an employee for his peer’s
achievement but rather under competitive contracts which reward an employee
Chair: Pinar Yildirim, University of Pennsylvania, Philadelphia, PA, for achievements better than his peer.
19104, United States, [email protected]
1 - Human Judgment and A.I. Pricing
Avi Goldfarb, University of Toronto, Rotman School of
Management, 105 St George Street, Toronto, ON, M5S 3E6,
n TA03
Canada, [email protected], Ajay Agrawal, Room 34, Alter Hall
Joshua Gans
Managing Information in Marketing
Recent artificial intelligence advances can be seen as improvements in prediction.
We examine how such predictions should be priced. We model two inputs into General Session
decisions: a prediction of the state and the payoff or utility from different actions Chair: Doug J. Chung, Harvard Business School, Boston, MA, 02163,
in that state. The payoff is unknown, and can only be learned through United States, [email protected]
experiencing a state. It is possible to learn that there is a dominant action across
all states, in which case the prediction has little value. Therefore, if predictions 1 - Does Restricting Information Make the Crowd More Accurate?
cannot be credibly contracted upfront, the seller cannot extract the full value, and Vineet Kumar, Yale University, 131 Dartmouth Street, 4th Floor,
instead charges the same price to all buyers. Boston, MA, 02116, United States, [email protected]
2 - The Editorial Power of Newspapers Crowdsourcing has been widely used for fairly accurate information gathering,
Markus Mobius, Microsoft Research, Cambridge, MA, 02142, and digital technologies enlarge the scope. We examine how herding in
United States, [email protected], Susan Athey, Jeno Pal, crowdsourcing can arise based on how much information is made available to
participants in such a crowd setting. Using a randomized experiment in
Sida Peng partnership with a financial information provider, we evaluate how users who
Newspaper have agenda setting power on their landing pages: a newspaper can receive different informational treatment respond and how overall quality and
prioritize articles in terms of (a) whether to cover a story at all, (b) how quantity of information can be enhanced.
prominently to display a story (in terms of horizontal and vertical placement as
well as font size). At the same time, readers of a newspaper have preferences over 2 - Using Deep Learning to Overcome Privacy and Scalability Issues
news articles which affect the order in which they consume news. We separately in Customer Data Transfer
estimate the agenda-setting power of newspapers as well as reader preferences Clarence Lee, Cornell University, 401b, Sage Hall, Ithaca, NY,
using a unique dataset of scraped US newspaper landing pages which record the 14853, United States, [email protected]
x/y position of each headline in 30 minute intervals. Disentangling these two
Customer privacy is becoming an increasingly important topic for marketers.
forces has many interesting applications. For example, a frequent criticism of
Recent high profile hacks of databases containing sensitive customer information,
aggregators and other news intermediaries is that they wrest editorial control
and the growing need to build data infrastructures sufficient to support analysis of
from newspapers and therefore reduce diversity of opinion. However, this
“big” data, present nontrivial obstacles to researchers seeking to utilize individual-
argument implicitly assumes that newspapers have significant agenda setting
level customer data from firms. In this paper, I show that recent developments in
power. Agenda setting power also raises the question of “optimal” agendas: are
machine learning may enable researchers to transfer models, instead of data, as a
newspaper agendas tailored to the preferences of the most opinionated readers or
potential alternative to the process of anonymizing and sampling customer data
towards the median reader? This questions matters in the light of increasing
for release. I demonstrate how effectively several types of deep learning models
media fragmentation which might increase news consumption polarization if
can preserve desired characteristics of original data, and compare estimation and
moderate readers read more biased news.
counterfactual results as estimated from real versus the synthetic customer data
3 - Information Shocks and Internet Silos: Evidence from Creationist generated by the transferred models.
Friendly Curriculum 3 - The Value of Firm-initiated and User-initiated Quality Signals:
Ananya Sen, MIT, Cambridge, MA, United States, An Empirical Analysis of Apartment Rental Platform
[email protected], Catherine Tucker Lingling Zhang, University of Maryland, College Park, MD, United
Studying factors that influence adoption of new products and technologies lies at States, [email protected], Fan Feng, Vithala R. Rao
the heart of marketing. In this paper, we study a manager’s decision to adopt
Information asymmetry is an acute problem faced by many service-oriented
automation for a production process. Automation offers efficiency by generating
platforms such as apartment rental platforms. Apartment renters (or consumers in
consistent high input at low operation costs. But at the same time, it reduces the
general) typically have less information about the quality of the apartment than
interaction among employees and reduces a manager’s ability to take advantage
the platform or the provider, causing inefficiency in matching. In this context,
of employees’ peer monitoring capacity. We study how automating some tasks in
there is hardly any repeated transactions and hence no opportunity for providers
a production process influences the effort of the employees working on the same
to develop their reputations. Without a reputation system, both platforms and
process production, and whether and when a manager should choose to
providers necessarily use various other means to signal quality. In this research,
automate the production process. Comparing the costs of an automated
we study the effect of platform-initiated and providerinitiated quality signals. We
production system to one that is not, we show that while automation provides
collaborate with a leading online apartment rental platform in an Asian market.
significant cost-cutting benefits, it may still result in an overall costlier system
Our dataset includes detailed records on more than 540,000 listings from a major
because of the increase in the cost of incentivizing the remaining employees. We
metropolitan city. The platform issues a quality badge for selected apartments,
show that as some tasks in production can be automated and automation
which we label as firm-initiated signals. Meanwhile, apartment owners (the
becomes gradually more cost-efficient, human teams are less likely to be
platform’s business users) provide text description to signal quality, which we
employed under cooperative contracts which reward an employee for his peer’s
label as user-initiated. We perform text mining to identify the most informative
achievement but rather under competitive contracts which reward an employee
features from the unstructured information provided by the business users. Given
for achievements better than his peer.
that the quality signals may be determined endogenously, we adopt the causal
4 - Men and Machine: When Should a Firm Adopt Automation? forest method (Wager & Athey 2017) with matching and form the identification
Mustafa Dogan, Carnegie Mellon University, Pittsburgh, PA, strategy by utilizing a large set of covariates. We focus on the heterogenous effect
United States, [email protected], Pinar Yildirim of quality signals and quantify how the effect varies by key characteristics of the
apartments. Our research has direct implications on how platforms and service
Studying factors that influence adoption of new products and technologies lies at providers with limited reputation information can influence consumer choices
the heart of marketing. In this paper, we study a manager’s decision to adopt through the disclosure of quality information.
automation for a production process. Automation offers efficiency by generating
consistent high input at low operation costs. But at the same time, it reduces the
interaction among employees and reduces a manager’s ability to take advantage
of employees’ peer monitoring capacity. We study how automating some tasks in
a production process influences the effort of the employees working on the same
2
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 3
4 - The Comprehensive Effects of Sales Force Compensation: scale panel discussion series for an exploratory pre-analysis. To estimate the
A Dynamic Structural Analysis of Performance and Selection model, we conducted a large-scale field experiment in cooperation with a major
Doug J. Chung, Harvard University, Morgan Hall fashion and sporting goods retailer. Results suggest that consumer engage in
mental trade-offs of privacy concerns and benefits - so-called privacy calculus.
161 Soldiers Field, Boston, MA, 02163, United States,
The relationship between privacy concerns and benefits from personalized
[email protected], Byungyeon Kim advertising seems to be mediated by the extent of customer’s imperfect
We provide a comprehensive model of salespeople’s behavior in response to information. By exploiting a large-scale quasi-experiment in cooperation with an
compensation plans. The model takes into account many of the key behavioral online retailer for furniture, interior design, and home textiles, study 2 quantifies
elements that constitute a realistic sales force setting; present-bias, forward the effect of opt-out on consumer behavior. The findings suggest that consumer
looking behavior, allocation of effort, and stay or leave decision of a salesperson. opt-out is economically meaningful. However, without more structure, it is not
Combined, the paper provides insights on how different components of possible to conclude what part of the effect was due to information asymmetries.
compensation affect performance and selection on heterogeneous salespeople. In To analyze this problem, we set up a structural model of advertising avoidance
addition, the paper introduces a key methodological contribution to the and use the additional restrictions that the model provides to estimate its
marketing literature. We provide a formal proof that distance-to-quota act as an parameters. Estimates are used for counterfactual simulations.
exclusion restriction that can be used to identify discount factors in an infinite
horizon setting. In addition, we identify conditions under which a quasi-
2 - Duration of Advertising Effects on Brand Choice: How do
hyperbolic discounting model is identified. Empirically, we find evidence of Recency Effects Change According to Time from Ad Exposure to
present-bias in salespeople’s behavior. Category Purchase?
Hiroshi Kumakura, Professor / Visiting Scholar, Chuo University /
New York University, 742-1 Higashi-Nakano, Hachioji, Tokyo, 192-
n TA05 0393, Japan, [email protected]
The principle of recency, closeness in time between advertising exposure and
Room 232, Alter Hall category purchase, has been proposed to discuss short-term and behavioral ad
Panel Discussion on Digital Marketing Applications of effects on brand choice in advertising literature. This principle asserts that a single
ad exposure near purchase occasion exerts a powerful influence on sales, and the
AI and Deep Learning I principle is useful for especially digital and mobile marketing focusing on the
consumer who is close to making a purchase decision. In recency advertising
Invited Session planning, although the timing of ad exposure is the most important and is
1 - Panel Discussion on Digital Marketing Applications of AI and attributed to the duration of ad effects, the duration by a single ad exposure has
Deep Learning I rarely been discussed. Hence, we discuss 1) how long advertising effects on brand
Moderator: Xueming Luo, Temple University, 1801 Liacouras choice by a single ad exposure continue, 2) how they are strong, and 3) how they
Walk, Philadelphia, PA, 19122, United States, [email protected] decline. Namely, we measure recency effects and show how they change
according as time passes, by estimating the probability densities of time (hours)
AI and DL algorithms will leverage big and rich data like sentiment, image, video, from consumer’s ad exposure to his/her category purchase, and by focusing the
and diverse subtleties of human interactions in one end-to-end process, not gap of shapes among these probability densities. Here, advertising effects of an
captured by traditional marketing. Empowered by deep neural networks, new individual consumer on an individual brand per outlets are to be discussed with
marketing applications in the industry are astonishing (academic works are hierarchical Bayes model. One of managerial implications of this research is that
lacking): smart chatbots with personality, consumer classification and unmet stronger recency effects are observed in a consumer having no loyal brand and in
needs detection, user preference discovery and recommendation, social semantics outlets of convenience stores and vending machines, in which a consumer may
and personalized content curation, digital fraud and data breaches prevention, use the product at or near a store soon after his/her purchase and the price is
intelligent customer service and ad targeting, omnichannel marketing and higher without price promotion, than in supermarkets.
dynamic pricing, and autonomous retailing and platform ecosystems.
3 - A Model Integrating the Multi-agent Simulation and State-space
Panelists Representation for Understanding the Interaction Effect of
John R Hauser, MIT, Sloan School of Management, Advertising and WOM
77 Massachusetts Ave., E62-538, Cambridge, MA, 02139, Eiji Motohashi, Associate Professor, Yokohama National University,
United States, [email protected] 79-4 Tokiwadai, Hodogaya-ku, Yokohama, 2408501, Japan,
Harikesh Nair, Assistant Professor of Marketing, [email protected], Sotaro Katsumata, Akihiro Nishimoto
Stanford University, Stanford, CA, United States,
This study aims to propose the model integrating the multi-agent simulation and
[email protected] state-space representation and understand the interaction effect of advertising
Xiao Liu, New York University, Stern School of Business, 44 W 4th and WOM. WOM has had important role in consumer’s purchase behavior, so
Street, New York, NY, 10013, United States, [email protected] companies have to consider the effect of WOM in order to implement the
effective advertising planning. We incorporate the dynamics among consumers
Sanjog Misra, University of Chicago Booth School,
and heterogeneity in purchase decision making to the model. The particle filter is
5807 S Woodlawn Ave, Suite 369, Chicago, IL, 60637, used to estimate the model. It is shown that the proposed model is beneficial to
United States, [email protected] predict sales and understand the consumer’s purchase decision making process.
4 - Understanding the Effects of Cross-media Consumption During
n TA06 Super Bowl Advertising
Prasad Naik, University of California, Graduate School of
Room 234, Alter Hall Management, Davis, CA, 95616, United States,
[email protected], Neeraj Bharadwaj, Michel Ballings
Advertising Effects
Consumers increasingly view a television program on a primary screen and attend
Contributed Session to social media on a second screen. During such cross-media consumption, how
Chair: Eiji Motohashi, Yokohama National University, 79-4 Tokiwadai, do stimuli from multiple screens influence viewers’ emotional response? To
Hodogaya-ku, Yokohama, 2408501, Japan, [email protected] address this open question, we construct a dataset over a two-year span, extract
factors from structured and unstructured variables, and estimate the effects of
1 - Consumer Privacy, Imperfect Information, and Marketing those factors from primary (i.e., television ads during the Super Bowl) and
Avoidance: Evidence from Retargeting secondary screens (i.e., consumers’ Facebook posts) on ad likability. Results show
Alex Eiting, TU Braunschweig, Schleinitzstrasse 23a, that both screens contribute to the explained variance. A novel aspect of this
Braunschweig, 38106, Germany, [email protected] study is to discover, based on the psycholinguistic literature, the three thinking
styles (viz., formal, analytic, narrative) using not only the “bag of words”
The most common method of identifying and tracking online consumers’ activity approach, but also the grammatical and syntactical role words play, thereby
involves the placement of cookies on a consumer’s hard drive that are then transcending the typical sentiment analysis and broadening our understanding of
offered back to the website during subsequent visits by the consumer. As online simultaneous media consumption.
advertising becomes more personalized, firms run the risk that tracking will elicit
a “creepy” feeling of being followed and that customers will find the advertising
invasive of their privacy and that reactance will lead them to resist the ad’s
appeal. Against this background, the upcoming General Data Protection
Regulation (GDPR) builds on the European Data Protection Directive 95/46/EC
with additional key requirements including tighter requirements for obtaining
valid consent to the processing of personal data. So any “do not track” mechanism
is assumed to facilitate consumer self-selection by their expected net benefit. To
learn about the motivations and mechanisms behind consumer opt-out and to set
up our econometric model, we accessed Federal Trade Commission’s fairly large-
3
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 4
n TA07 n TA08
Room 237, Alter Hall Room 238, Alter Hall
Retailing I CB - Unfamiliarity & Choice
Contributed Session Contributed Session
Chair: Yoonju Han, Kelley School of Business, Indiana University, Chair: Paul Wang, University of Technology Sydney, Marketing
1309 E. 10th Street, Hodge Hall 2100, Bloomington, IN, 47405, Discipline Group, P.O. Box 123, Sydney, NSW 2007, Australia,
United States, [email protected] [email protected]
1 - The Effect of Competition on Retail Price and Variety Decisions: 1 - Consumer Preference for Opaque Products
Theory and Evidence Lucas Stich, Ludwig Maximilians Universität München,
Fabio Caldieraro, Associate Professor of Marketing, FGV - EBABE, München, Germany, [email protected], Martin Spann, Gerald Häubl
Rio de Janeiro, R. Jornalista Orlando Dantas, 30, Rio de Janeiro, Selling opaque products refers to a marketing practice where sellers offer goods or
22231-010, Brazil, [email protected], Andre G. Trindade services for which they conceal some product attributes from buyers and reveal
The effect of competition on retail price and variety decisions: theory and them only after a non-refundable purchase. Examples include selling hotel rooms
evidence We study the impact of competition on a firm’s pricing and product on Priceline and Hotwire, lucky bags in retailing (e.g., Fukubukuro; surprise
variety decisions both theoretically and empirically. We focus on grocery stores, boxes on Amazon), or surprise menus in restaurants. Prior research is analytical
an industry where variety is particularly important to consumers. By looking at a in nature, has focused primarily on the sellers of opaque products and explained
large number of markets, we find that both prices and variety are higher when consumers’ choice of opaque products by a lower price to compensate for the
there are two competing stores than in those markets with a single store. We uncertain product characteristics. In contrast, we propose that consumers can
propose and test a model that explains these patterns based on two main forces: have a preference for the unique aspects of opaque products and not necessarily
(a) incentives to increase variety increase with competition because of a business need a discount to choose an opaque alternative when non-opaque products are
stealing effect; (b) a larger set of products allows firms to better screen also available. We propose choice difficulty, exploration, and excitement as
heterogeneous consumers leading to a strategic price-increasing effect under candidate pathways to explain consumers’ choice of opaque products, which we
stronger competition. We also find that consumer welfare is higher under test in a series of experiments. We find that consumers choose opaque products
competition, not because of prices, but because of the net increase in consumer more frequently the higher the perceived difficulty of making a choice. In
utility due to the better (wider) availability of products in the market. addition, both the desire to discover unfamiliar things as well as the anticipated
pleasure from experiencing surprise (and uncertainty resolution) can induce
2 - Measuring Product Differentiation using Market consumers to choose opaque products. Our results also indicate that opaque
Share Distributions products can be an effective means to motivate consumers to make a choice
Young Han Bae, Assistant Professor of Marketing, Penn State rather than delaying it.
University, Greater Allegheny, 4000 University Drive, Frable 217, 2 - A Unified Model of Context-dependent Preference
McKeesport, PA, 15132, United States, [email protected], Prithwiraj Mukherjee, Assistant Professor, Indian Institute of
Hyunwoo Lim, Thomas S. Gruca, Gary J. Russell Management Bangalore, Bannerghatta Road, Bangalore, 560076,
It is well known that the relationship between the size of market share and India, [email protected], Arnaud De Bruyn, Ayse Onculer
market share rank follows a power law. Specifically, the market share power law
We present a generalized theoretical model of context-dependent preferences
implies that a log-log plot of market share size versus market share rank is linear
based on a multi-attribute utility theory formulation, where we assume that
with a negative slope. The most likely explanation for this type of relationship is
attribute-wise utilities are stable, but weights on each of these is shaped by
that brands in a product category have true quality values drawn from the
context, including choice set, past experience, word of mouth and advertising, via
extreme upper tail of a product quality distribution. Using this fact, we argue that
an anchoring mechanism. Our theoretical model explains and proposes boundary
slope of the log-log plot provides a measure of the amount of quality
conditions to several documented biases including the compromise effect,
differentiation in a product category. Moreover, by compiling a dataset of log-log
attraction effect and status quo bias. Using this parsimonious theoretical model,
plot slopes for many categories, it is possible to study how product category
we also derive a choice model and compare it to existing models using choice-
characteristics impact product differentiation. We implement this approach using
based conjoint data. We also present data from some additional studies to test the
a dataset containing the market shares of 5,733 brands in 562 consumer packaged
robustness of our approach. Our model has several implications for marketing
goods categories in the United States. We first show that power law market share
managers, in product line design, new product design and product differentiation.
distributions are a valid empirical generalization. We then show that the
parameters of the power law distributions show considerable variation across 3 - Does the Information Influence Choice of New Healthy Food
product categories, and that this variation is significantly related to category-level in Males and Females? – Role of Neophobic Trait and – Type of
measures of marketing activity. Our empirical results provide strong evidence that Unfamiliarity
strong product differentiation is associated with categories that are perceived to be
Varisha Rehman, Assistant Professor, Indian Institute of
important by consumers and that have lower levels of price promotions and
couponing. We discuss implications of our approach both for marketing managers Technology Madras, Sardar Patel Road, Adyar, Chennai, 600036,
and researchers in marketing strategy. India, [email protected], Sujatha Manohar
3 - The Dynamics of Shopper Engagement and Purchase Conversion Due to the alarming health issues around the world, there is an urgent need for
the food marketers and policy makers to inculcate healthy food choices among
at Retail Checkout
the people.The objective of the research was twofold, firstly to study the impact of
Yoonju Han, Indiana University, 1309 E. 10th Street, Hodge Hall information (health and taste) on willingness to try new healthy foods and
2100, Bloomington, IN, 47405, United States, [email protected], secondly the role of type of unfamiliarity of the food, food neophobia and gender
Raymond R Burke, Shibo Li, Alex Leykin in influencing healthy food choices. A pseudo-experiment was conducted with
The checkout area or “front end” of a supermarket sells an ever expanding 385 subjects with 12 versions of the survey where the new dish and the
assortment of products, and is one of the most productive areas of the store and information presented were manipulated. Results show that the gender and type
an important driver of trip satisfaction and future store loyalty. Prior research on of unfamiliarity have a direct effect on willingness but the information does not
retail checkout has focused on how waiting time affects customer satisfaction and have. This could be because the subjects were mostly young and they are more
the implications for queue management. Little is known, however, about how influenced by social and environmental factors rather than by information. The 3-
shoppers attend to, engage with, and shop products while waiting in line. This way interaction effect of type of unfamiliarity with the information is significant
research proposes a simultaneous equations model which predicts that, as only for females which suggest that males ignore the information. Since food
shoppers move through the checkout lane, their motives, abilities and contextual neophobia is found to be significant on willingness for both males and females,
factors dynamically influence their eye fixations, product interactions, and marketers should try to reduce their reluctance. The study differentiates
purchases. We estimate the model using a unique dataset of eye-tracking and unfamiliarity and novelty in the context of food. The results would help the food
survey data collected from a field experiment at a large U.S. grocery chain. The marketers to provide the right kind of information in influencing the people to
research reveals that product engagement and “impulse purchases” are the choose a new healthy food. It could also help in deciding on the factors such as
predictable outcome of interactions between personal, contextual and marketing packaging, labelling etc. and in designing their advertisements according to the
factors. The study provides marketing researchers and store managers with type of unfamiliarity of the food.
valuable insights for product organization, merchandising, and line management
at retail checkout.
4
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 5
4 - Association Between Decision States and Media Preference in a 3 - You Can Skip this Ad in 5 Seconds: Exploring Consumers’
New Consumer Tech Purchase Ad Skipping Behavioron Internet
Paul Wang, University of Technology Sydney, Marketing Aditya Billore, Associate Professor, IIM Indore, Prabandh Shikhar,
Discipline Group, P.O. Box 123, Sydney, NSW 2007, Australia, Rau-Pithampur Road, Indore, 453556, India, [email protected],
[email protected], David Waller, Mark Morrison, Manoj Motiani
Harmen Oppewal Consumers encounter advertisements while browsing websites, these ads appear
Consumer decision states are discrete behavioral states that correspond with their in-page before, during, or after streaming in any web content. Consumers show
particular levels of knowledge, information search, and readiness to make a their rejection or appreciation by skipping the ad or sharing them with other
purchase. Consumer motivation for the information search would vary depending users using social networking sites. The current study attempts to understand
on their decision state. Research on decision states is especially important for both how the consumers decide to skip or watch and share the ad. We propose to use
marketing researchers and practitioners to better understand high involvement need for cognitive closure as the theoretical foundation for explaining the ad
purchase decisions. Although many studies have observed the use of the Internet skipping behavior. The proposed experiment will explore the ad skipping
as an important vehicle for purchasing, relatively little research attention has behavior in context of video advertisement on video content delivering websites
been paid to how consumers use multiple media channels such as shop displays, (e.g. YouTube).The major constructs used to explore ad skipping behavior are
advertising, and sales assistants to search for information and make their purchase Consumer perceived Advertising Creativity (Divergence and Relevance), Brand
decisions. To fill the research gap, we conducted an online survey of more than Familiarity and ad Avoidance. In this study first we try to understand how ad
1400 respondents to empirically investigate the association between decision skipping is different from ad avoidance. The later part of the study tries to explore
states and media preference in a new consumer tech purchase. In addition, the how brand familiarity and perceived ad creativity influence the decision of
study examines the effect of demographics on consumer media preference. The consumer to skip or watch the ad. Contrary to the prevalent view regarding the
study has found that consumers use of information varies across decision states, ad avoidance is that consumers are likely to process and appreciate ads from
gender, and age. The findings suggest that it is often inadequate and erroneous to known brands as compared to unknown brand. We here propose that in case of
treat consumers in different decision states as a homogeneous market. ad skipping the consumers tend not to skip ads from unknown brand because of
novelty as unfamiliar brands may reduce the need for cognitive closure (NCC).
The results of the study will help in understanding the determinants and
consequences of ad skipping behaviour. How ad skipping is different from ad
n TA09 avoidance? And what is the role of brand familiarity consumers’ decision about
ad skipping.
Room 239, Alter Hall
4 - Music Matters: Do Audio Features of Accompanying Music Affect
Understanding Ads Advertising Effectiveness?
Contributed Session Joonhuyk Yang, Doctoral Student, Northwestern University,
Evanston, IL, 60208, United States,
Chair: Aditya Billore, IIM Indore, Prabandh Shikhar, Rau-Pithampur
[email protected], Purushottam Papatla,
Road, Indore, 453556, India, [email protected]
Lakshman Krishnamurthi, Caiyun Liu
1 - The Effect of Mobile Search Ads Across Devices:
A Geo Experiment The literature suggests that creative aspects play a significant role in how
consumers respond to advertising. Specifically, findings indicate that the creative
Michelle Andrews, Assistant Professor of Marketing, Emory
features of an ad affect its likeability which in turn influences purchase intentions
University, 1300 Clifton Rd, Atlanta, GA, 30329, United States, for the advertised product. In light of the above findings, there have been calls for
[email protected], Ting Li, Francesco Balocco increased research into the creative dimensions of advertising - one of which is
Understanding the role of mobile search ads in the path to purchase is critical for accompanying music. There is however virtually no research into how the
managing multiple-device campaigns. Since sales attributions to search origins characteristics of accompanying music affect consumer response. The goal of our
across devices often hinges on individual- or device-level identification, we research is to address this gap and develop insights into the audio characteristics
overcome the problem of multiple-identifiers by designing a field experiment in of accompanying music that makes advertising more effective. Our investigation
which we geo-split consumers into treatment and control groups. Modeling the proceeds in two stages. In stage one, we use two large datasets that provide
effect of treatments between regions enables us to detect lifts in key advertising information on the popularity of music along with its audio characteristics. The
performance metrics on tablets and desktops in the same regions where mobile first, the Million Song Dataset, provides data on the number of times that over a
search ads received the experimental treatment. We then investigate the effect of million users listened to more than 380 thousand songs. The second, Free Music
mobile search ads on same and cross-device performance by varying ad bids on Archive, provides similar data for over 106 thousand songs. Both datasets also
mobile keywords based on their position in the purchase funnel. Our multi- provide data on audio characteristics like rhythm, tone and tempo of the songs.
period geo-experimental evidence comprises more than 3,000 ad campaigns Our goal for this stage is to identify audio characteristics that increase the number
served to mobile, tablet and desktop that generated more than 70 million of times that people listen to songs while controlling for factors like genre, release
impressions, more than 2 million clicks, and more than 75,000 conversions. Our date, artist popularity and duration. In stage two, we assess whether our findings
results highlight device interdependencies by demonstrating the nuanced manner carryover to the context of advertising by investigating the relationship between
through which the position of mobile keywords in the purchase funnel may advertising effectiveness and the characteristics of accompanying music.
heterogeneously drive same and cross-device conversion and shape the
information gathering process in the shopping journey.
2 - When is the Sweetest Time for TV Commercial?
Chihiro Totsuka, Graduate Student, Keio University, Felica Hiyoshi
#202 3-5-12 Kohoku Yokohama, Kanagawa, 223-0061, Japan,
[email protected]
TV commercials still have greatest impact on consumers in Japan and a lot of
companies invest large portion of their advertising budget in TV commercial.
However, it is difficult to know the effect of TV commercials and each company
blindly uses TV commercials in their own ways. In this research, I tried to find out
the most effective time zone for TV commercials and to know why the effect
become highest at a certain time zone. The reason why I focus on time zone is
that it is comparatively easy for companies to control time zone when they place
TV commercials. The goal of this research is to propose the way how each
company can make the most of their TV commercials. I used a single source data
collected by Nomura research institute, which involves data of how much people
come in daily contact with TV and websites, and data concerning awareness,
preference and purchase experience towards products and I compare the scores of
consumers’ intention and purchase towards products before they come in contact
with TV commercials and after that. Moreover, using data about consumers’
demographics and usages of web sites and SNS, I analyzed the reason why some
time zones have greater effect on consumers deeply.
5
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 6
n TA10
benefiting appeals will consider a difficult to process campaign to be less favorable
than an easy to process campaign. We tested a moderated mediation using a sense
of guilt to explain the psychological mechanism. In Study 1, 311 US participants
Room 605, Alter Hall from Amazon Mechanical Turk were randomly assigned to one of four conditions:
Consumer Judgement & Choice 2 (appeal: social vs. self) × 2 (processing fluency: easy vs. difficult). Participants
were presented with an appeal scenario (White & Peloza 2009) and a prosocial
Contributed Session campaign to process. The level of lexical processing fluency was manipulated as
Chair: Vlada Pleshcheva, Humboldt University Berlin, being easy or difficult to understand (Zhang & Mattila, 2015). As predicted,
Wirtschaftswissenschaftliche Fakultÿt, Institut für Marketing, Berlin, people rated that they are less likely to volunteer for disfluent campaign with self-
benefit appeal (F(1, 305) = 4.30, p = .039) through marginally increasing feeling
10099, Germany, [email protected] of guilt. In Study 2, 197 US participants were randomly assigned to one of four
1 - Cascade Effect in Service Quality Evaluation conditions: 2 (appeal) × 2 (processing fluency), and we used the same procedures
Sheila Roy, Associate Professor, S.P. Jain Institute of Management in Study 1 by changing the sharpness of text and pictures in the campaign
and Research, I 704 Raheja Vistas Raheja Vihar, Chandivalli, (Labroo & Kim, 2009). Further, as predicted, we found that guilt mediates the
Mumbai, 400072, India, [email protected], Indrajit Mukherjee interaction effect on attitudes (95% CI = [.011, .371]).
We attempt to extend the theory on evaluative judgements of partitioned 4 - Do Consumers Value Identical Improvements in Fuel
(sequential) experiences (Ariely and Zauberman, 2003) and the theory on Consumption and Co2 Emissions of Cars Equally?
preferences for sequences of outcomes (Lowenstein and Prelec,1994), to service Vlada Pleshcheva, Research Assistant, Humboldt University Berlin,
quality evaluations across multiple stages in a service process. Prior research in Wirtschaftswissenschaftliche Fakultät, Institut für Marketing,
partitioning of sequential experiences suggest that the overall evaluation of the Berlin, 10099, Germany, [email protected]
experience is impacted by summary evaluations of each partition rather than
gestalt characteristics such as peak experiences, trend or end. In choice literature, The present study investigates how the framing of information on the
when choices are framed as sequences of outcomes, people tend to prefer environmental impact of vehicles affects consumers’ preferences and willingness-
improving sequences and delayed gratification. The above research seems to to-pay (WTP) for identical improvements in a car quality. In online choice
assume that experience of prior stages in a sequence would not impact either the experiments, the effects from two metrics (fuel consumption vs. CO2 emissions)
evaluation or choice of future stages. However, in a service setting which is and three scales of one metric (CO2 in 0.100 kg/km vs. 100 g/km vs. 10,000
usually designed as a series of stages partitioned by logical breaks, the experience g/100 km) are examined. First, from a technical view, fuel consumption (FC) and
of prior stages may impact evaluation or choice of subsequent stages and the CO2 emissions are linearly connected by a constant factor and are thus
overall evaluation of the sequence may be mediated by evaluation of isomorphic in describing the environmental friendliness of a car. Second, rescaling
intermediate stages. This phenomena is referred to as a cascade effect in identical information should not change consumer decisions. However, as this
operations management literature (Shi and Zhou, 2009; Sulek et al., 2006). The study demonstrates, consumers’ perception of and WTP for identical
objective of the research is to study the cascade effect of prior stage evaluation on improvements in the car quality vary significantly depending on the framing of
subsequent stage customer experience and on overall evaluation of the information. The study’s contribution lies in comparing effects from two measures
experience. Through an experimental design of a multistage education loan of the same information, not within one measure (FC or CO2) as in previous
application process we study the mediating cascade effect of intermediate stages research, and in contrasting the investigated metric and scale effects among
on final service quality evaluation. The findings of the study have managerial consumers who prefer either diesel or gasoline vehicles. The estimation accounts
implications for the design of customer experiences. for heterogeneity in tastes, environmental attitudes, knowledge, and importance
of climate change issues for the respondents. The insights of this study serve to
2 - When are Comparison Sites Used in Consumer Search provide guidance for both policy makers and car manufacturers on how to
Peng Yam Koh, PhD Candidate, Singapore Management present information on car offers.
University, 50 Stamford Road, Singapore, 178899, Singapore,
[email protected], Ernst Osinga,
Peter Sander van Eck
n TA11
Comparison sites allow consumers to compare product and service offerings from
different providers on price and non-price characteristics and are ubiquitous in Room 606, Alter Hall
consumer search. While search theory can explain why consumers search on
comparison sites, it cannot explain when consumers search on these sites.
Digital Marketing I
Moreover, this theory provides little guidance for firms who wish to list on Contributed Session
comparison sites. To fill this gap, we develop theory relating key constructs from
search theory to the timing of comparison site visits in the consumer’s search (i.e., Chair: Gabriela Alves Werb, Goethe-Universität –
earlier or later in the search). We test our theory using: (i) household panel data Wirtschaftswissenschaften, Theodor-W.-Adorno-Platz 4, Frankfurt am
capturing online browsing, purchase behavior, and pre-search survey on the Main, 60596, Germany, [email protected]
Dutch health insurance market, and (ii) a Bayesian probit model with 1 - Quantifying the Impact of Interactive Marketing Feature:
endogenous search equation to address possible self-selection by consumers. We Evidence from a “Daily Check-in” Experiment
find that the moment of comparison site usage depends on the consumer’s: (i)
Tianfu Wang, Purdue University, 3384 Peppermill Dr, Apt 2A,
size of the initial consideration set, (ii) dissatisfaction with their current product,
(iii) expectation of finding a better deal, and (iv) switching intention. Overall, we West Lafayette, IN, 47906, United States, [email protected],
deepen search theory by explaining when and for what reasons consumers Xing Fang
consult comparison sites. Our findings have important managerial implications for Online retailers traditionally attract customer traffic via advertising, price
managers of comparison sites and managers of providers listing on comparison discount, and other promotional events. In recent years, new interactive
sites. marketing features, such as “like” (e.g., Facebook), “check-in” (e.g., Yelp), and
3 - Illusion of Processing Fluency on Pro Social Campaigns “follow” (e.g., eBay) have emerged. Despite the tremendous marketing resources
devoted to incentivize such behaviors, little is known about their effectiveness,
Unjustifiable Efforts Produce Guilty Feelings and if so, what is the mechanism behind them? We utilize a quasi-experiment to
Yaeeun Kim, Temple University, Philadelphia, PA, United States, answer this question. We collaborate with an online retailer to design and
[email protected], Yaeri Kim, Vinod Venkatraman, Kiwan Park implement a week-long daily check-in event. The customers who visit and
Previous studies have revealed contradictory effects of processing fluency on “check-in” with the retailer receive loyalty points as a reward each day. The
attitudes. When processing advertisements, easy to process stimuli are more customers who participate in the daily check-in event serve as the treatment
positively evaluated as a feelings-as-information (Reber et al., 2004), while group while the customers who do not participate in the event serve as the
difficult to process stimuli are more favorable as an effortful investment (Deval et control group. We compare both the short-term and long-term changes in
al., 2012; Labroo & Kim, 2009). It is still not clear if the evaluation of disfluent purchase behavior across the two groups of customers. In the follow-up study, we
advertisements is motivated with a desirable appeal (e.g., social-benefiting appeal) try to resolve the customer self-selection issue via a randomized field experiment.
or an undesirable appeal (e.g., self-benefitting appeal). In the current research, we These results provide the first evidence of the cost-effectiveness of “daily check-
predict that such perceived disgraceful actions increase unpleasant emotional in” and explore the mechanism of interactive of marketing features.
states such as guilt, which would affect their attitudes (Kouchaki et al., 2014;
Peloza et al., 2013). We explore this effect by applying disfluent stimuli to a
prosocial behavior campaign. Thus, we hypothesize that people primed with self-
6
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 7
2 - Morphing for Consumer Dynamics 2 - When Does More Information Mean Less for Consumers
Alina Ferecatu, Assistant Professor of Marketing, Rotterdam Xiaoyan Xu, PhD Candidate, National University of Singapore,
School of Management, Erasmus University, Burgemeester 21 Lower Kent Ridge Road, Singapore, 119077, Singapore,
Oudlaan 50, Rotterdam, 3062 PA, Netherlands, [email protected], [email protected], Yuetao Gao, Wei-Shi Lim
Gui Liberali In some service industries (e.g., credit cards, mobile phone plans, insurance),
Website morphing automatically changes the “look and feel” of websites to match many consumers do not carry the service plan that best fits their needs. Recently,
the cognitive styles of its visitors. Current developments in morphing theory service-switching infomediaries have entered the market and consumers can
assume that consumers’ cognitive styles, and therefore their information needs locate better service plans by using websites’ free service if the infomediaries can
and preferences, are constant over time. However, changes in their information reach them. An infomediary is an Internet-based information provider that
needs can impact the effectiveness of a morph over time. We propose an gathers information about consumers’ preferences based on their existing
algorithm that allows for learning under dynamic information preferences. We consumption behaviors and recommends service providers to them; however, the
account for dynamic decision-making stages, such as when website visitors move choices are restricted to only the service providers enrolled with the infomediary.
from the information-gathering stage to the consideration stage, and into the We examine how provision of information by an infomediary affects service
purchase stage. We characterize the learning of the match between morphs and providers’ profits and consumer surplus. We find that when a large proportion of
decision-making stages as a multi-armed bandit problem, which we solve with a the uninformed consumers are not matched with their ideal service, the
dynamic allocation index. We develop a novel dynamic programming algorithm infomediary profits by adopting a non-exclusive contract to enroll all service
to explicitly model the within-visit trade-off between nudging a consumer providers. In addition, enrolling with the infomediary is a dominant strategy for
through the purchase stages, and risking having her terminate the website visit. the service providers. However, they may consequently realize less profit.
We design morphs based on categorization theory, which provides us with Furthermore, we show that although the infomediary intends to provide
substantive support for the choice of morph content and language tailored for information such that consumers can make a better service choice, consumers are
early and late stages of the purchase funnel. We report a longitudinal field worse off if the reach of the infomediary to uninformed consumers is high.
experiment to establish a proof-of-concept for our method. Consumers are better off if the reach is low.
3 - Visibility-at-risk: An Approach to Estimate a Firm’s Risk of Losing 3 - Analyzing Lost Customer Behavior in the Business Analytics Era
Visibility in Organic Search Ke Li, Assistant Professor, New Jersey City University, 110 First St,
Gabriela Alves Werb, Goethe-University Frankfurt, Theodor W. Apt 20J, Jerse City, NJ, 07302, United States, [email protected],
Adorno-Platz 4, RuW-Building, Room 1.236, Frankfurt am Main, Anthony C. DiBenedetto, Eric M. Eisenstein
60323, Germany, [email protected], Winning back lost customers has gained more and more importance in businesses
Christian Doppler, Bernd Skiera since winning back lost customers can recapture lost revenue and at the same
Organic search clicks already constitute the most important source of online time costs much less than acquiring new customers. The extant papers on
traffic for firms in several industries, such that a loss of visibility in organic search winback mainly focused on developing strategies to regain customers that were
may severely affect a firm’s ability to generate profits. We propose an approach to lost during the retention stage only and examine factors such as customer
estimate a firm’s risk of losing visibility in organic search results, with the aim to perceptions of fairness, customers’ first-lifetime satisfaction through survey, price
provide insights into its size, heterogeneity and economic consequences. Half of discounts, service-based win-back offers, and defection reasons on the likelihood
the firms in our sample of more than 1,000 firms face a 5% probability of losing of reacquiring these lost customers. In this paper, we explore how customers’
more than 55% of their visibility within one year. This result suggests that many experiences, customer engagement and company’s touch point in customers’
firms that rely on organic search clicks to drive traffic to their domains have a journey during the conversion period and retention period respectively influence
considerable risk associated with their future profits. Our estimates also reveal customer reacquisition probability within the competitive environment. Different
that the risk of losing visibility is highly heterogeneous across industries. A machine learning methods that could be used for winback segmentation are also
backtesting analysis with out-of-sample data suggests that our developed risk discussed. Our study helps firms to understand how to apply different winback
measure, Visibility-at-Risk, is a valid measure for the risk of losing visibility in strategies to different segments of their lost customers in the business analytics
organic search results. Finally, our results outline that changes to a firm’s visibility era.
index have significant and prolonged effects on profits. These results have
important implications for managers and investors.
n TA13
n TA12 Room 746, Alter Hall
Multi – Channel Strategies I
Room 745, Alter Hall
Contributed Session
Marketing Strategy - Markets & Analyses
Chair: Penelope Schoutteet,Vrije Universiteit Brussel, Pleinlaan 2,
Contributed Session Brussels, 1050, Belgium, [email protected]
Chair: Ke Li, New Jersey City University, 110 First St, Apt 20J, 1 - Dominant Retailer, Lower Product Quality and
Jersey City, NJ, 07302, United States, [email protected] Uninformed Customers
1 - Optimal Couponing with Required Purchase Quantity Axel G. Stock, University of Central Florida, Dept of Marketing
Aharon Hibshoosh, Professor, San Jose State University and CBA, P.O. Box 161400, Orlando, FL, 32816-1400, United States,
Lincoln University, 401 15th Street, Apt 10, Oakland, CA, 94709, [email protected], Somnath Banerjee
United States, [email protected], Uri Ben Zion, Uriel Spiegel A number of consumer and business reports suggest that slightly lower quality (or
The model describes a common deal routinely offered by Kroger and others. The feature) variants of products are being sold through dominant and mass retailers,
model is an extension of related coupon rationing models by Ben Zion, Hibshoosh while higher quality variants continue to be sold through specialty and weaker
and Spiegel (e.g. BHS 2000). A profit maximizing firm is offering a coupon with a retailers and, customers are uninformed about such differences. We study two
stipulation that every consumer using the coupon must purchase C items of intriguing questions based on this phenomenon viz. (1.) why are lower quality-
product x. The firm simultaneously optimizes its decision variables: the coupon lower priced variants sold primarily through dominant retailers and not the
discount level D and its required coupon purchase quantity C, under nominal weaker retailers? (2.) why do the weaker retailers or manufacturers not inform
rigidity. The profit’s C and D elasticities are denoted as eC and eD, and the optimal customers about these quality differences? Using a game theoretic model, we find
values as C* and D*, correspondingly. We adhere to our couponing framework that (1.) in a bilateral monopoly of manufacturer and retailer an increase in
where in contrast with the restricted parametric modeling in marketing we bargaining power of the retailer leads to a lower product quality in the channel, if
assume a general market response function of the decision variables. quality is non-contractible. Consumer surplus rises, while social welfare follows
Furthermore, we model the consumer choice in a commodity space rather than in an inverted-u shaped curve with increase in retailer dominance and the
some spatial framework with separable consumer utilities. As common in associated decrease in product quality. (2.) The weaker retailer does not have an
couponing modeling we assume marketing to two market segments: a top full incentive to communicate its higher quality even if the advertising is costless, if
price loyal market segment and a lower deal prone one, assuming downward the quality differences are not too high. The motivation for this result is traced to
leakage. We identify parsimonious functions of elasticities which, in general, the threat of increased competition in the event of such advertising.
determine the optimal values. For example, we prove that: a) at optimum eD >
eC ; b) D* depends only on the gap eD-eC. The optimal coupon shares of
segments are simple fractions involving only linear functions of eC and eD. Based
on the general results we also develop as private cases parametric results and
generate explicit solutions for C*and D*, a rarity in the couponing literature.
7
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 8
8
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 9
9
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 10
3 - Technological Outsourcing May Not Fill the Gap: Investigating 2 - Shipping Fees and Product Assortment in Online Retail
Reactive Licensing in the Pharmaceutical Industry Donald Ngwe, Harvard Business School, Boston, MA, 02163,
Manuel Ignacio Hermosilla, Johns Hopkins University, United States, [email protected], Chaoqun Chen
100 International Dr, Baltimore, MD, 21201, United States, We seek to measure how consumers respond to value-contingent free shipping
[email protected] incentives and how online retailers can use this information to design optimal
Technological innovators routinely strengthen their new product pipelines by shipping policies that work together with pricing and product assortment. Using
outsourcing developing products from other companies. We argue that this historical sales data from an online fashion retailer, we document model-free
strategy may backfire if carried out reactively. That is, if the firm uses outsourcing evidence of purchase patterns around minimum order requirements for free
to fill the gap left by the failure of other products in its pipeline. In these cases, shipping. Orders tend to bunch just over the free shipping minimum order value,
urgency may make the firm more vulnerable to the asymmetric information with consumers “topping up” purchase baskets by choosing higher priced or
problems that plague markets for developing technologies. Supporting evidence additional items. We build a demand model that characterizes these purchase
from the pharmaceutical industry is presented. Outsourcing is operationalized by patterns in an environment with multiple product categories. This model enables
drug candidate licensing contracts in this context. Exploiting quasi-experimental us to measure sensitivity to product prices and shipping fees separately. We use
variation in pipeline gaps, we show that: (i) large pharmaceutical firms engage in this model to simulate demand under counterfactual scenarios, including
reactive licensing, (ii) relative to candidates licensed proactively, those licensed alternative shipping fee schedules and product assortment decisions. We find that
reactively are more likely to fail post-licensing development, and (iii) financial firms can substantially improve performance by jointly determining product
markets do not “distinguish” reactive from proactive licensing on a systematic attributes and shipping policies.
basis. 3 - Scarcity Rents in Car Retailing: Evidence from Inventory
4 - Service Performance and Customer Reallocation: Customer Fluctuations at Dealerships
Urgency, Rationing, and Cherry Picking Ayelet Israeli, Harvard Business School, Morgan Hall 177, Soldiers
Tae Jung Yoon, University College London, 604 Duckman Tower, Field Road, Boston, MA, 02163, United States, [email protected],
3 Lincoln Plaza, London, E14 9BL, United Kingdom, Florian Zettelmeyer, Fiona Scott Morton, Jorge Silva-Risso
[email protected] Price variation for identical cars at the same dealership is commonly assumed to
In this paper, I investigate three underlying mechanisms about why quality arise because dealers with market power are able to price discriminate among
information disclosure may not be beneficial to customers in the service industry their customers. In this paper we show that while price discrimination may be
where sellers are often capacity-constrained. First, customers who are urgent and one element of price variation, price variation also arises from inventory
cannot wait their turns may be matched to low quality sellers. Second, if prices fluctuations. Inventory fluctuations create scarcity rents for cars that are in short
are regulated in the market, proactive seller rationing can reallocate unprofitable supply. The price variation due to inventory fluctuations thus functions to
customers to low quality sellers because high quality sellers often face higher efficiently allocate particular cars that are in restricted supply to those customers
demand after information disclosure and thus can strategically seek out profitable who value them most highly. Using transaction level data on new car sales across
consumers. Third, if disclosed quality measures depend on customer the US between 1998 and 2014, we find that a dealership moving from a
characteristics, sellers can cherry-pick customers who do not ruin their quality situation of inventory shortage to an average inventory level lowers transaction
measures. Using an exogenous policy shock to disclose healthcare providers’ prices by about 0.5% ceteris paribus. Shorter resupply times also decrease
performance, I estimate structural models and disentangle these three transaction prices for cars in high demand. For traditional dealerships, inventory
mechanisms. Then I find that quality information disclosure can decrease explains 50% of the combined inventory and demographic components of the
customer welfare mainly because less urgent customers choose high quality predicted price. For so-called “no-haggle’’ dealerships, the percentage explained
sellers ahead of more urgent customers. by inventory increases to 62%.
4 - Using Observational Data to Increase Accuracy in
Marketing Experiments
n TB02 Ron Berman, The Wharton School, University of Pennsylvania,
700 Jon M. Huntsman Hall, Philadelphia, PA, 19104-6340,
Room 33, Alter Hall United States, [email protected], Elea McDonnell Feit
Digital Economy II: Customer Interactions and Recent research analyzing advertising experiments has shown that simple
Product Delivery randomized controlled trials designed to measure the average treatment effect
across all ad viewers are often underpowered and that unreasonably large
General Session samples are required to prove that a campaign is profitable. In this paper we show
Chair: Ron Berman, The Wharton School, Philadelphia, PA, that if the advertiser can rank viewers based on their responsiveness to
advertising, then the power to detect advertising response can be substantially
19104-6340, United States, [email protected]
improved by stratifying viewers into high- and low-responsiveness groups and
1 - Balancing Service Speed and Agent Utilization for an Online using a post-stratified estimator of the advertising effect. Analytically, we show
Insurance Platform that past observational data about consumers can be used to rank consumers by
Andres I. Musalem, University of Chile, Beauchef 851, Santiago, their treatment effects, even if the past exposures were not-random. Thus firms
8370456, Chile, [email protected], Marcelo Olivares, with prior panel data on customer response to advertising (e.g. CRM data) can
Daniel Yung use this observational data to increase the power of their holdout experiments.
We apply the stratification approach to re-analyze a direct mail experiment
In this research, we study an online platform that sells car insurance policies conducted at large specialty retailer. We estimate prior responsiveness for each
offered by different companies. After a customer visits the platform’s website and customer based on previous CRM data describing catalog mailings, emails and
enters vehicle, demographic and contact information, the customer is shown transactions for individual customers and then use this information to stratify
several car insurance quotes. Customers who do not purchase online are then customers. In several experiments, the stratified analysis improves the accuracy of
contacted via telephone by the platform’s call center agents. Using data from this the estimator enough to determine that a campaign was effective when it was
platform and estimating econometric models with instrumental variables we find previously thought otherwise.
that as the time that it takes the platform to contact a customer increases, the
probability that the customer purchases one of the insurance policies decreases.
This suggests that the platform might gain from improving the speed at which it
contacts its website visitors. One approach to accomplish this goal is to increase
the call center capacity by hiring more agents. However, doing so reduces their
utilization and may hence reduce the compensation obtained by each agent. We
empirically find that a reduction in an agent utilization translates a higher
probability of the agent quitting her job, thus reducing the average experience of
the call center agents. Since more experienced agents exhibit a better
performance at converting quotes into purchases, this creates a tradeoff for the
platform where on the one hand a greater call center capacity leads to better
speed of contact but also reduces the incentives for experienced agents to keep
their jobs at the platform. In this paper, we empirically quantify this tradeoff and
provide guidelines in terms of how these capacity decisions should be made.
10
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 11
n TB03
goods purchased. The effect of salience on quality accounts for at least one-third
of the overall revenue decline.
Room 34, Alter Hall 3 - A Simple and Robust Estimator for Discount Factors in Optimal
Marketing-Operations Interface Stopping Dynamic Discrete Choice Models
Oeystein Daljord, University of Chicago, Chicago, IL, United States,
General Session [email protected], Denis Nekipelov,
Chair: Jiong Sun, Purdue University, West Lafayette, IN, 47907, Minjung Park
United States, [email protected] We propose a simple two-step estimator for the discount factor in a class of
Co-Chair: Juncai Jiang, University of Texas-Dallas, Dallas, TX, 75252, dynamic discrete choice models. The estimator follows from a constructive
identification result and has an intuitive interpretation: it recovers the discount
United States, [email protected] factor as the sensitivity of current period choice probabilities to shifts in the
1 - The Impact of Wal-mart’s Sales Information Disclosure on continuation values. The estimator is derived as the solution to a single well-
Supplier Performance behaved moment condition which is linear in the discount factor and is
Juncai Jiang, Virginia Tech, Blacksburg, VA, United States, independent of the utility function. The estimator is therefore easy to implement,
[email protected], Chenxi Zhou, Jiong Sun computationally light, and in contrast to existing estimators, robust to biases from
finite sample approximations to the unknown utility function. We apply the
This paper investigates whether and how the stoppage of a dominant retailer’s estimator to data on mortgage defaults under an identifying assumption of time
sales information disclosure affects supplier performance. Specifically, we focus on homogeneity of the utility function. We compare the performance of the
the monthly Comparable Store Sales (CSS) index that represents the retail sales proposed estimator to alternative two-step estimators that jointly estimate the
growth rate for stores that have been opened up for at least one year, and discount factor and the utility function. The results show that our proposed
empirically investigate the change in suppliers’ financial performance when Wal- estimator’s robustness to finite sample approximation bias and its computational
Mart withdraws monthly CSS. We found that the cessation of Wal-Mart’s ease do not necessarily come at material expense of precision.
monthly CSS disclosure is associated with negative supplier abnormal returns.
However, not all suppliers are hurt equally: the negative financial response is 4 - Retail Competition with Online Shopping
more pronounced for suppliers with higher sales volatility and suppliers who Yufeng Huang, University of Rochester, Rochester, NY, 14627,
served Wal-Mart only (compared with those who supplied to both Wal-Mart and United States, [email protected],
Target Inc.). Finally, the withdrawal of Wal-Mart monthly CSS is also associated Bart Bronnenberg
with higher supplier inventory volatility for those who served Wal-Mart only.
How does online shopping impact spatial competition in the retail industry? Using
2 - Partial Vertical Ownership in the Presence of large individual-level panel data from the Dutch retail apparel industry, this paper
Downstream Competition directly measures the diversion ratio between retailers in their brick-and-mortar
Jiong Sun, Purdue University, 812 W State St, West Lafayette, IN, and online channels across a variety of retail formats. We find that online
47907, United States, [email protected], Fang Fang, shopping captures high-value customers, intensifies competition between retailers
Baojun Jiang but only through expanding consumer choice sets, and is complementary to the
brick-and-mortar channel within a retail chain.
Partial vertical ownership plays an important role in aligning the incentives of
firms involved in vertical relationships. This paper examines the impacts of partial
vertical ownership on pricing decisions, firm profitability, and consumer surplus.
We show that intense competition may hurt consumers when the acquiring firm n TB05
is a low-value firm.
Room 232, Alter Hall
3 - Return Policy and Product Design in Direct and Indirect
Distribution Channels Panel Discussion on Digital Marketing Applications
Buqing Ma, University of Science and Technology of China, of AI and Deep Learning II
Hefei, China, [email protected], Lu Hsiao, Yunchuan Lin Invited Session
We study firms’ return policy and product quality design in direct vs. indirect Chair: Jing Peng, University of Connecticut, Storrs, CT,
distribution channels. We find that contrary to conventional wisdom, sellers may
[email protected], Raghuram Iyengar, Kartik Hosanagar
offer less lenient return policy in direct distribution channels and we also
characterize the market and cost conditions under which either channel offers 1 - Panel Discussion on Digital Marketing Applications of AI and
misaligned product quality and return policy. Deep Learning II
Moderator: Xueming Luo, Temple University, 1801 Liacouras
Walk, Philadelphia, PA, 19122, United States, [email protected]
n TB04 AI and DL algorithms will leverage big and rich data like sentiment, image, video,
Room 35, Alter Hall and diverse subtleties of human interactions in one end-to-end process, not
captured by traditional marketing. Empowered by deep neural networks, new
Research on Consumer and Firm Behavior marketing applications in the industry are astonishing (academic works are
lacking): smart chatbots with personality, consumer classification and unmet
General Session needs detection, user preference discovery and recommendation, social semantics
Chair: Yufeng Huang, University of Rochester, Rochester, NY, 14627, and personalized content curation, digital fraud and data breaches prevention,
United States, [email protected] intelligent customer service and ad targeting, omnichannel marketing and
dynamic pricing, and autonomous retailing and platform ecosystems.
1 - Market Segmentation and Managerial Effort
Kosuke Uetake, Yale School of Management, 165 Whitney Panelists
Avenue, Room 5477, New Haven, CT, 06520, United States, Catherine Tucker, MIT, 1 Amherst Street, E40-167, Cambridge,
[email protected], Kevin Williams MA, 02142, United States, [email protected]
We document significant heterogeneity in pricing strategies both within and Anindya Ghose, New York University, New York University,
across firms using novel retail panel data. Some categories exhibit follow the New York, NY, 10012, United States, [email protected]
leader strategies but there is also considerable variation in probability of response Puneet Manchanda, University of Michigan, Ross School of
and response time across products and categories. We present a model of Business, Ann Arbor, MI, 48109-1234, United States,
behavioral firms where price adjustments are costly and firms are subject to
[email protected]
inattention. We use a moment inequalities estimator to bound managerial costs.
With the model estimates, we decompose the role of inattention and effort in Carl F Mela, Duke University, Fuqua School of Business, Box
rationalizing observed price adjustments. Finally, our counterfactual simulations 90120, Durham, NC, 27708-0120, United States, [email protected]
study the occurrence of uniform pricing and quantify its welfare implications in
oligopoly.
2 - Price Salience and Product Choice
Sarah Moshary, University of Pennsylvania, Philadelphia, PA,
United States, [email protected], Tom Blake,
Kane Sweeney, Steven Tadelis
We study the effect of price salience on product choice along two dimensions:
whether a good is purchased and, conditional on purchase, the kind of good
purchased. Consistent with our theoretical predictions, we find that making the
full purchase price salient to consumers reduces both the quality and quantity of
11
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 12
n TB06
multi-source attribution contexts. The proposed model is an extension of the
popular proportional hazards model, in the sense that it allows the cause of an
event to be multiple sources. Our model has several advantages as compared to
Room 234, Alter Hall past modeling efforts in multi-source attribution. First, the model does not
Advertising and Effectiveness speculate on the contribution of each channel apriori, but lets the data
automatically determine the contribution of individual channel based on its
Contributed Session characteristics. Second, it is easy to use and interpret as compared to structural-
Chair: Jing Peng, University of Connecticut, Storrs, CT, form models for attribution, and more rigorous as compared to previous
[email protected], Raghuram Iyengar, Kartik Hosanagar rule-based attribution methods. Third, it provides unbiased estimates even if only
part of the sources contributes to the conversion of consumers. We demonstrate
1 - Exploring Whether the Lifting of the Hedge Fund Advertising Ban the effectiveness of our model using both simulated and real-world datasets. Our
has Resulted in Smarter Investments model has implications for marketers in several areas, especially multi-channel
advertising and social media marketing.
Navid Mojir, Harvard Business School, Morgan Hall 141,
Soldiers Field, Boston, MA, 02163, United States, [email protected],
Andrew Sinclair
The ban on advertising in the hedge fund industry, going back to the Securities n TB07
Act of 1933, was recently lifted as part of the Jumpstart Our Business Startups Act Room 237, Alter Hall
(i.e. the JOBS Act) of 2012. While opponents of the JOBS Act argue that poor-
performing hedge funds would take advantage of it by targeting less-sophisticated Retailing II
investors, its proponents argue that in a market largely considered as a B2B
market - with most of the investors being large sophisticated institutions -
Contributed Session
allowing advertising would reduce search cost for investors and improve the Chair: Byungkee Min, PhD Student, Syracuse University,
allocation of funds in the market as a whole. To assess these arguments, we 721 University Avenue, Syracuse, NY, 13244, United States,
combine data on hedge funds from multiple sources, including Lipper TASS [email protected]
database, Form ADV from SEC, and the Wall Street Journal. Taking advantage of
the fact that the JOBS Act affects only US-based investors, we use a difference-in- 1 - A Hidden Side of Consumer Grocery Shopping Choice
difference strategy to explore the effects of the JOBS Act. We find that: i) Hedge Aidin Namin, Assistant Professor of Marketing, Loyola Marymount
funds in fact do respond to the lift of the advertising ban by reducing their use of University, One LMU Drive, MS 8395, Los Angeles, CA, 90045,
“traditional” publicity measures. ii) The JOBS Act has resulted in an increase in United States, [email protected], Yashar Dehdashti
fund flows to hedge funds. Interestingly, the media is playing a more important
role in increasing fund flows to hedge funds after the JOBS Act. iii) The lift of the One of the important decisions which grocery stores’ managers deal with, mostly
advertising ban has resulted in investors paying more attention to more salient on a daily basis, is exercising better control over store traffic on different days of
but misleading predictors of performance when making investment decisions (i.e. the week. This issue may, then, be reflected in better handling of pricing,
performance-chasing). promotions, display, and layout decisions in the store. This study analyzes and
identifies hidden classes of grocery store customers and their choice of items
2 - Automation of Dynamic Marketing Attribution Models purchased on a grocery store trip on different days of the week. Following the
Harpreet Singh, Kvantum Inc., 5188 Selma Avenue, Fremont, CA, literature, the three major groups of grocery shopping products are investigated in
94536, United States, [email protected], this research; these are food/drink, cleaning, and personal care. Using a rich U.S.
Prasad Naik, Shilpi Sharma grocery store scanner dataset, commonly purchased pair of products from each
group of grocery products are selected and examined. Employing an advanced
Marketing performance tracking and optimization can be modeled as a non- latent class analysis technique, i.e., Finite Mixture Modeling, latent classes (i.e.,
linear, dynamic problem solved through ensemble Kalman filters. The key segments) of customers, their sizes, and customers’ choice of grocery items on
challenge in a practical marketing mix setting is to be able to initialize the Kalman different days of the week are revealed and empirically validated. The developed
filter. The use of BFGS or EM type of algorithms does not provide robust solutions model controls for consumer unobserved heterogeneity and accounts for
when estimating the model’s variance parameters. Applying MCMC techniques inclusion of demographic variables which guide the mixing probabilities. Major
combined with Kalman filters, we have developed a robust reliable method that findings indicate that there are two largely different-in-size latent classes for the
avoids manual tinkering of model estimation, thereby enabling significant food/drink, two relatively similar-in-size for the cleaning, and three different-in-
automation. This method can be further extended and made computationally size for the personal care group. For each latent class, descriptive demographics
robust through multiple banks of filter ensembles. Our team has successfully and buyers’ choices on different days of the week are clearly identified. Unveiling
deployed these robust innovations in real world applications for multiple CPG & such hidden patterns has meaningful managerial implications and sheds light on
Retail companies globally. We will present our methods and applications that marketing mix, store traffic management, pricing, and promotion managerial
significantly reduce computational times, thereby empowering brand managers decisions made by retail managers.
towards agile decision-making.
2 - Consumer Trust and Online Payment Options: Determinants of
3 - How Effective is Product Placement as a Marketing Tool? E-commerce in the Least Developed Countries
Simha Mummalaneni, University of Washington, Foster School of Mohamed Muse Hassan, Student, Ritsumeikan Asia Pacific
Business, University of Washington, Seattle, WA, 98195-3226, University, Minamitateishi, Ikku 5-5, Yamaguchi Apartment, No.
United States, [email protected], Pradeep Chintagunta, 1, Room 720, Beppu, 874-0839, Japan, [email protected]
Sanjay K Dhar, Yantao Wang
The behavior of consumers in the least developed countries (LDCs) is different
Product placement provides a way for brands to reach consumers in a more subtle than the behavior of consumers in USA or Europe, mainly due to the cultural,
way than through traditional advertising. We use data from both traditional social, and economic contexts (Akman & Rehan, 2014). Therefore, this study
advertising and product placement on television shows to compare how both investigates the impact of consumer trust and online payment options available
approaches affect consumer demand for brands in the soda, diet soda, and coffee on the awareness level and perception of e-commerce, and propensity to shop
categories. Our approach is to estimate a logit demand model using store-level online among consumers in the least developed countries, taking Somalia as a
sales data, while accounting for heterogeneity in consumer preferences and case study since the country has achieved the second highest Internet growth in
response parameters across markets. Estimates from this model indicate that Africa after Democratic Republic of Congo between 2000 and 2017 (ITU, 2017).
product placement is generally effective, but that the elasticities are small: average We developed a model with five constructs, and empirically tested the model with
short-term elasticities are around 0.01 for the major brands in the data, which is 500 respondents. The paper employs a survey design approach, and 500
lower than the estimated elasticities for traditional TV advertising but on the same questionnaires were distributed with 78 percent completion rate. The main
order of magnitude. However, there is a significant amount of heterogeneity in findings of the study indicate that the support of mobile payment option by the e-
elasticities across categories, brands, and geographical markets; in the coffee retailers is a critical factor in Somalia as one-third of the population has mobile
category and for Diet Coca-Cola, we find that product placement is more effective bank accounts, compared to less than 10% who use traditional bank methods.
than traditional TV ads. The study also found that awareness of e-shopping is very low, and Somali people
have mixed reactions toward the perception of online shopping as this is still a
4 - A Novel Approach to Attribution in Decisions Influenced by new endeavor in its infancy stage. Instead, Somalis prefer cash on delivery (COD)
Multiple Sources method. The main conclusion is that Somalis would embrace e-commerce if a
Jing Peng, University of Connecticut, 2100 Hillside Road, mobile option is allowed as part of the payment options available. The paper
Unit 1041, BUSN 368, Storrs, CT, 06269, United States, concludes with a discussion of the managerial implications of the study.
[email protected], Raghuram Iyengar, Kartik Hosanagar
Consumers often make decisions under the influence of multiple sources (e.g.,
promotions, ads, and friends) and it’s important for managers to identify the
contribution of each source in the decision process. Despite the obvious
managerial implications of the multi-source attribution problem in consumers’
decision-making process, there is limited past research primarily due to the lack of
a suitable methodological framework. In this paper, we propose a general
reduced-form regression model that can be easily applied to various kinds of
12
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 13
3 - Optimal Return and Re-sale Policies for E-tailers is here that education is currently poised to generate the most positive externality
Byungkee Min, PhD Student, Syracuse University, 721 University for the society. The goal and contribution of our study is to address this
Avenue, Syracuse, NY, 13244, United States, [email protected], conspicuous research gap in the literature. Specifically, using the World Bank’s
data from a statistically representative survey of nearly14,000 urban households
Eunkyu Lee
in Kenya, our study sheds systematic insights into the key determinants of
Due to the limited observability of product and uncertainty of product fit with household decisions on whether and how much to spend on education. We
consumer preference, e-tailers are encountering a large number of product present novel findings into the roles of social identity, financial leverage, and
returns. Combined with the widespread lenient return policies, the skyrocketing digital access. Of particular interest are the roles played by the educational
returns volume is triggering significant cost issues for e-tailers. Meanwhile, some attainments of parents as well as the genders of children within a household. We
customers have started to complain that e-tailers are selling previously returned discuss how and why do these roles occur, and how they can inform the broader
goods as new products, even when the consumers can detect it is not truly new issues faced by families, providers, and regulators.
due to imperfect packaging. In a profit-maximizing firm’s perspective, it can
decide to re-sell non-defective returns. However, e-tailers should keep in mind 3 - Adoption Intention of Kidney Patients for E-health Device –
that such re-selling practice might affect the online consumer behavior and Mediating Effect of Social Capital
produce adverse effects for the firm. The purpose of this study is to identify the Fang Chi Chang National Chung Hsing University, Taichung City,
optimal return and re-sale policy for e-tailers when re-selling returned goods is Taiwan, [email protected], Ming Chih Tsai, Shiau-Chi Lin,
allowed. This paper develops an analytical model taking into account the current Yi Shin Chen
e-commerce environment, and seeks to identify optimal strategies for better
product returns management. Our initial results indicate a significant impact of Internet of things (IoT) has increasingly become new solutions across sectors. As
category-specific market environment and inventory salvage value on the firm’s the technology incessantly evolved, customer adoption becomes essential for
optimal choice of pricing and return policies. success. In essence, IoT involves multiple-players in operations. Social issue is
significant in affecting technology adoption, but it is yet well discussed in extant
innovation studies. By integrating theories of social institution, social capital and
technology acceptance model, this study develops an analytical model for
n TB08 assessing patient behavior of adopting e-health device. Taiwanese kidney patients
(totally 3.2 million, ranked as 1st in terms of medical resource consumption)
Room 238, Alter Hall adopting e-self-heath control devices providing on-line health information for
CB - Social Influences interaction with medical physician are targeted for empirical study. The model
consists of 7 constructs investigating the social influences of kidney patient on
Contributed Session adopting the e-health device, where social capital between patient and physician
(p/p) is treated as a mediating variable and treatment length as a moderating
Chair: Qin Zhang, Pacific Lutheran University, 12180 Park Avenue variable. Totally 221 valid samples classified into two treatment length groups are
South, Tacoma, WA, 98447, United States, [email protected] collected through face-to- face interview with the aids of a teaching hospital in
1 - Contagion of the Competitive Spirit: The Influence of a central Taiwan. The result indicates that the adoption behaviors of the kidney
Competition on Non-competitors patients vary by treatment length. For the long-time patients, the e-health care
may result in excessive social capital and thus reduce the adoption intention.
Vincent Mak, Cambridge Judge Business School, University of
Conversely the new patients would adopt the e-device as with the great benefit of
Cambridge, Trumpington Street, Cambridge, CB2 1AG, United increasing the p/p social capital. The study result provides important information
Kingdom, [email protected], Raghabendra KC, Marcus Kunter helping hospital improve p/p social capital and the IoT developers justify
From customer contests to innovation awards, competitions are ever-present in marketing strategy.
social life. The recent rise in gamification strategies, in areas such as marketing 4 - The Neighborhood Effect in Charitable Giving
and crowdsourcing, further popularizes attempts to motivate people by engaging
them in competitions. However, such initiatives may not always induce full or
Qin Zhang, Assistant Professor of Marketing, Pacific Lutheran
majority participation among the target population: it is typical in a competition University, 12180 Park Avenue South, Tacoma, WA, 98447,
that the competing individuals are far outnumbered by people who do not United States, [email protected], Sang-Uk Jung
participate in it but are aware of it. We report a series of experimental studies that We propose a spatial autoregressive model to investigate the neighborhood effect
investigate the influence of a competition on non-competitors who do not on consumers’ charity giving behavior. The empirical study uses a dataset that
participate in it but are aware of it. Our first study is a large-scale field experiment contains donation information of 946 consumers to a single charity organization
involving pay-what-you-want entrance at a German zoo (N = 22,886). We find over five years in the North Island of New Zealand. The results show a significant
that customers who were aware of a competition over entrance payments, but did positive neighborhood effect on the amount of donation. This indicates that
not participate in it, paid more than customers who were unaware of the consumers’ donation behaviors to a charity are interdependent and their
competition - establishing the presence of the contagion effect. The second study decisions of how much to donate are influenced by others in their social network.
finds further confirmatory evidence for the effect in a controlled laboratory This finding has implications on marketing strategies for non-profit organizations
setting. The third study eliminates alternative explanations and necessitates the and suggests that a geographically concentrated approach is likely to solicit more
need for competition for the contagion effect to occur. The fourth experiment donation than direct effort on individual consumers.
provides further confirmatory and process evidence for the effect, showing that it
is driven by heightened social comparison motivation due to mere awareness of
the competition. Moreover, we find evidence that the reward level for the
competitors could moderate the contagion effect on the non-competitors. Even if n TB09
an individual does not participate in a competition, their behavior can still be
influenced by it; and this influence can change with the characteristics of the Room 239, Alter Hall
competition in an intriguing way.
Personalization, Engagement & Choice
2 - Roles of Social Identity, Financial Leverage, and Digital Access on Contributed Session
Households’ Spending Behaviors on Education: Insights From a
Sub-saharan Country Chair: James Dearden, Lehigh University, Bethlehem, PA,
Debu Talukdar, Professor, University at Buffalo, 215 E Jacobs United States, [email protected]
Management Center, Buffalo, NY, 14260, United States, 1 - A Personalization Field Test From Industry
[email protected] Richard Loa, Canadian Broadcasting Corporation,
The critical role of education in our contemporary society can hardly be 250 Front St W,, Toronto, ON, Canada, [email protected]
exaggerated. For an individual, education is often the key to better employment The Canadian Broadcasting Corporation (CBC) is performing an experiment to
and economic security. From an aggregate perspective, educational attainments of prolong session extension while delivering utility to the Canadian. The
its people have huge positive externality effects on the quality of a country’s mechanism for delivering the utility is providing a list of reading
human capital and thus on its long-term economic development. Not surprisingly, recommendations on the CBC search page independent of search engine results.
worldwide annual spending underlying consumption of educational services runs In this session we will briefly touch upon the motivation of the experiment,
into trillions of dollars. Given its pivotal social and economic role, research experiment design, results of the experiment and next steps the CBC plans to take
insights into factors affecting spending on education are of keen interest to both after knowing what they know now.
education providers and policy makers. However, data availabilities have meant
that research has mostly focused on public spending by governments. In contrast,
rigorous research on private spending behavior by households remains quite
limited. In fact, it is practically non-existent for African countries, even though it
13
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 14
4 - The More the Merrier? A Study of Knowledge Quality on an Online Prosper.com, an online crowd-funding platform. The authors find that loan
Knowledge Sharing Platform requests written by defaulting borrowers are more likely to include words related
Tingting Fan, The Business School of The Chinese University of to their family, mentions of god, short-term focused words, the borrower’s
financial and general hardship, and pleading lenders for help. The authors further
Hong Konog, Room 1113, 11/F, Cheng Yu Tung Building,
observe that defaulting loan requests are often written in a manner consistent
No. 12 Chak Cheung Street, Shatin, Hong Kong, with the writing style of extroverts and liars.
[email protected], Jia Liu, Leilei Gao
3 - Can User Generated Content Predict Restaurant Survival:
Online knowledge sharing (e.g., Quora, Zhihu, StackOverflow) has become a
trending way for people to seek knowledge from each other on internet.
Deep Learning of Yelp Photos and Reviews
Compared with traditional knowledge sharing organizations (e.g., schools, Lan Luo, University of Southern California, 3660 Trousdale
institutions), online knowledge sharing provides an open platform which allows Parkway, ACC306, Los Angeles, CA, 90089, United States,
knowledge to be shared and diffused in a bigger scale with a lower cost. Millions [email protected], Mengxia Zhang
of people have contributed a huge amount of knowledge to those platforms in With widespread usage of smartphones, 3 billion photos are shared on the
merely a few years and many online knowledge sharing platforms have been Internet daily. Nevertheless, very few papers have examined whether and how
valued billions of US dollars. Despite the dramatic growth of online knowledge the prosperity of a business may be affected by consumer posted photos. We use
platforms, they are all facing an urging challenge: how to incentivize people to deep learning methods to analyze 795,175 photos and 1,015,825 reviews posted
contribute high quality knowledge? Our research taps into this challenge by on Yelp from 2004 to 2015 on 17,796 restaurants. Tracking the survival of these
studying knowledge sharing behavior on Zhihu.com, one of the largest online restaurants during this time period, we find that both volume and valence of
knowledge sharing platforms in China. By collecting a panel of 132,000 users’ photos are strong predictors of restaurant survival. Nevertheless, when it comes to
shared knowledge (approximately 2 million answers on 1 million questions) since reviews, only valence (not volume) matters. Interestingly, even after controlling
the first day of their registration, we investigate (1) how does the quality of for the content, star rating, and length of reviews, consumer sentiment extracted
shared knowledge evolve over time? (2) how is the quality of a user’s shared from review text is still strongly associated with the survival of restaurants. We
knowledge influenced by other users’ shared knowledge? (3) how is the quality also find that chain restaurants and restaurants from larger categories have
of a user’s shared knowledge influenced by social interactions among other users? greater chance to survive. Ceteris paribus, price levels do not appear to impact
Using a panel data model with individual user fixed effects, we find that in the restaurant survival. Our research is among the earliest attempts to introduce both
first few days after a question is post online, a user contributes higher quality photo- and text- based deep learning in marketing. We are also the first to
knowledge if other users have shared a good amount of knowledge, but compare and contrast managerial impacts of consumer posted photos vs. reviews.
afterwards as the amount of shared knowledge increases, he is likely to contribute To our knowledge, this is also the first large-scale empirical research on restaurant
lower quality of knowledge. We also find that the quality of shared knowledge survival.
increases as the social interactions among other users increases.
4 - Targeting and Privacy in Mobile Advertising
Hema Yoganarasimhan, University of Washington, Seattle, WA,
n TB11 United States, [email protected], Omid Rafieian
Mobile in-app advertising is growing in popularity. While these ads have excellent
Room 606, Alter Hall user-tracking properties through mobile device IDs, they have raised concerns
Machine Learning Applications in Marketing: II among privacy advocates. There is an ongoing debate on the value of different
types of mobile targeting, the incentives of ad-networks to engage in behavioral
General Session targeting and share user-data with advertisers, and the role of regulation. To
answer these questions, we propose a modeling framework that consists of two
Chair: Hema Yoganarasimhan, University of Washington, Seattle, WA, components — a machine learning framework for predicting click-through rate
98195, United States, [email protected] and a stylized analytical framework for conducting data-sharing counterfactuals.
1 - Measuring the Impact of Trial Period Lengths using Causal Using data from the leading in-app ad-network of an Asian country, we show that
Forests: Evidence from a Field Experiment our machine-learning model improves targeting ability by 17.95% over no
Ebrahim Barzegary, University of Washington, 481 Paccar Hall, targeting. These gains mainly stem from behavioral information and the value of
contextual information is relatively small. Stricter regulations on user tracking
Foster School of Business, Seattle, WA, 98195, United States,
substantially shrink the value of behavioral targeting. Counterfactuals show that
[email protected], Hema Yoganarasimhan the total advertisers’ surplus grows with more granular information sharing
A commonly used customer acquisition strategy in the software industry is to between the ad-network and advertisers. However, there is heterogeneity among
offer consumers a limited period of time to try the product for free. However, no advertisers in their preferred level of data sharing. Importantly, the ad-network’s
research has examined the issue of optimizing the length of these free-trials. revenues are non-monotonic, i.e., it prefers to not share behavioral information
Using data from a large-scale field experiment run by a leading software firm, we with advertisers. Thus, the ad-network may have natural incentives to preserve
study the relative effectiveness of three trial lengths — 7, 14, and 30 days. We users’ privacy without external regulation.
find that the shortest trial length of 7 days is the best option at the aggregate level.
Next, we use the newly developed causal random forests methodology to uncover
the rich heterogeneity in treatment effects and quantify effects at the individual-
level. We then use our estimates to examine the returns to counterfactual
treatments. When we move all consumers to the 7-day free trial, we see a 4.42%
increase in subscriptions and an 8.11% increase in revenue/user. As we increase
the customization of the treatment, these numbers increase, e.g., when we assign
the best average trial-period for the country to all users from that country, the
subscriptions increases by 4.9% and revenue/user by 9.06%. When we fully
personalize trial-length at the user-level, subscriptions increase by 24.18% and
revenues/user by 41%. Our results emphasize the value of combining machine-
learning methods for causal inference with field-experiments to personalize
marketing mix variables.
2 - When Words Sweat: Identifying Signals for Loan Default in the
Text of Loan Applications
Alain Lemaire, Columbia University, New York, NY, 1027,
United States, [email protected], Oded Netzer,
Michal Herzenstein
The authors present empirical evidence that borrowers, consciously or not, leave
traces of their intentions, circumstances, and personality traits in the text they
write when applying for a loan. This textual information has a substantial and
significant ability to predict whether borrowers will pay back the loan over and
beyond the financial and demographic variables commonly used in models
predicting default. The authors use text-mining and machine-learning tools to
automatically process and analyze the raw text in over 18,000 loan requests from
15
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 16
16
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 17
17
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 18
n TB15
n TB16
Room 603, Alter Hall Room 231, Alter Hall
Brand Extensions and Image 2018 ISMS/MSI Gary Lilien Rehearsals
Contributed Session Rehearsal Session
Chair: Yewon Kim, University of Chicago, Booth School of Business,
Chicago, IL, United States, [email protected] n TB17
1 - Dimensions of Brand Extension Fit Room 31, Alter Hall
Paul R. Messinger, University of Alberta, Faculty of Business,
3-20e Faculty Of Business Bldg, Edmonton, AB, T6G 2R6, Canada, ISMS Doctoral Dissertation Proposal
[email protected], Qian (Claire) Deng Competition Winners
A sizable literature on brand extensions argues that the congruity between a General Session
parent brand and an extension category has a positive effect on consumer
reception of the extension. But the application of this literature is limited because Chair: Olivier Toubia, Columbia Business School, New York, NY, 10027,
of a lack of understanding of what “brand-extension fit” really is. The current United States, [email protected]
paper develops a measurement scale of Brand Extension Fit (the BEF scale) 1 - Letting Logos Speak: A Machine Learning Approach for
consisting of two core underlying dimensions, engineering-based and market- Data-Driven Logo Design
based congruity, each formed by three items. Our proposed scale represents a Ryan Dew, Columbia University, 521 W 112th Street, Apt 62, New
synthesis and extension of the various formative items of congruity used in
York, NY, 10025, United States, [email protected], Asim
papers in the literature that consider multi-dimensional measures of fit. We
validate the model with two separate datasets obtained from general consumers Ansari, Olivier Toubia
consisting of judgments about extensions for (a) fictitious parent brands and (b) Logos serve a fundamental role in branding as the visual figurehead of the brand.
real parent brands, providing excellent sample-resample reliability. We provide Yet, due to the difficulty of using unstructured image data, prior research on logo
various analyses using SEM (structural equation modeling) techniques to provide design has been largely limited to non-quantitative studies. In this work, we
evidence of nomological validity. Unlike uni-dimensional measures of fit - such as explore logo design from a data-driven perspective. In particular, we aim to
similarity, fit, consistency, or “makes sense” - used in most of the brand-extension answer several key questions: first, to what degree can logos represent a brand’s
literature, our scale provide guidance for opportunity identification, idea personality? Second, what are the key visual elements in logos that elicit brand
generation, the building of a marketing plan around a chosen alterative, and and firm relevant associations, such as brand personality traits? Finally, given text
understanding the pros and cons of various alternatives. describing a firm, can we suggest features of a logo that elicit the firm’s desired
brand identity? To answer these questions, we develop a novel logo tokenization
2 - On the Dynamics of Brand Extensions: The Case of Movies algorithm, that uses modern image processing and computer vision tools to
Sridhar Moorthy, Manny Rotman Chair in Marketing, University decompose unstructured pixel-level image data into theory-based, human-
of Toronto, 105 St Georges Street, Toronto, ON, M5S 3E6, Canada, meaningful, visual features. We then apply this tokenization algorithm in
[email protected], Masakazu Ishihara conjunction with methods from the machine learning literatures on
In movies, brand extension takes the form of sequels. This paper compares the dimensionality reduction and transfer learning, to a dataset of hundreds of logos.
sales dynamics of movie originals and their sequels. We find that sequels diffuse In our unique dataset, the logos are matched with textual descriptions from firms’
faster than originals—their sales are more front-loaded. This difference in websites, third party descriptions of the firms, and consumer evaluations of brand
diffusion speed is moderated by several movie characteristics. Our econometric personality, which represent many distinct views of the brand: visual, textual, and
methods take into account the endogeneity of the sequeling decision: only consumer-level. By learning mappings across these distinct views, our approach
successful movies spawn sequels, not all movies. We interpret these findings in uncovers links that exist between a brand’s logo, description, and personality, and
light of what they say about the different “information processes” underlying thereby helps us understand the underpinnings of good design, and inform the
original brands and their extensions. design of new logos.
3 - Proposal Method to Evaluate Brand Image by using 2 - Automation, Decision Making and B2B Pricing
Psychophysical Function Yael Karlinsky Shichor, Columbia Business School, New York, NY,
Masao Ueda, Waseda University, 3-4-1 Okubo, Shinjuku-ku, United States, [email protected], Oded Netzer
Tokyo, 169-8555, Japan, [email protected] In the past century, automation has changed the labor market by consistently
replacing repetitive and predictable human tasks. More recently, automation has
In order to maintain a brand power, brand managers want consumers to have a
tapped into tasks that require judgment and soft skills and involve business
good brand image in consumers’ memory for a long time. Nowadays, it is difficult
acumen. We investigate the potential impact of automating experts’ decisions in
for consumers to keep a particular brand image, since brand image is more
one of the most complex and knowledge intensive jobs in marketing - sales. We
affected by social media. To keep a good brand image, the brand managers should
explore the tradeoff between the possible improvement in decisions making as a
examine consumers’ brand image if necessary. For this purpose, a new research
result of automation and the loss of soft skills in the context of salespeople
method is required to equip easiness of implementation and to have enough
making pricing decisions in a business to business (B2B) environment. The B2B
amount of information for decision making. It is needed to satisfy this
market is estimated at $8 trillion in transactions and is largely lagging behind the
requirement, in the present study, in addition to the reaction to the item, the
B2C market in adoption of information technology and automation.
reaction time was also collected and the obtained data was analyzed using a
Consequently, and in light of increasing demand from business client for digital
model applying a function of psychophysics. In the empirical analysis, the author
business channels, the potential for automation is immense. Using secondary data
investigated the brand image of a certain automobile manufacturer A in Japan
from a B2B retailer We find that a bootstrap pricing model of the salesperson
and analyzed obtained data from about 1,000 respondents. It is assumed that the
generates expected profits equal to those of the salesperson, despite the loss of
model had random effects for the intensity of each brand image. The random
noncodeable information available to the salesperson but not to the model.
effect parameters in this model was estimated by Markov Chain Monte Carlo
Furthermore, we find that a hybrid pricing policy that follows the model’s prices
method and estimated results indicated that the proposed method is effective for
in most cases, but the salesperson’s prices in cases where crucial non-codeable
understanding brand images.
information exists, leads to a significant increase in profits to the firm. We then
4 - Natural Experiment on Co-branding: The Case of Museum conduct a high-stakes field experiment with the B2B retailer and find that
Membership Demand providing salespeople with their own model’s prices, in real time, via the CRM
Yewon Kim, University of Chicago Booth School of Business, system, increases gross profit to the company by over 10%, which translates to an
increase in profits of $40k over the eight days of the experiment and a projected
5807 S. Woodlawn Avenue, Chicago, IL, 60637, United States, increase of approximately $2.2 million in yearly profits. We find that the hybrid
[email protected], Sanjog Misra, Bradley Shapiro model should use the salesperson and her expertise in pricing complex orders
We study the effect of co-branding on consumer demand in the market for (e.g., special processing is required or the order has many lines) and orders by
museum memberships in California. Over the course of our sample, one major infrequent clients, and that automation is particularly beneficial for low-expertise
museum with a highly recognized brand closed. During the closure, it salespeople. Overall, our findings suggest, that automation of experts’ pricing
sequentially co-branded with two established local museums. With individual decisions in B2B settings is not only feasible but can also lead to increased
panel data on museum memberships, we measure the effect of co-branding on profitability. We show empirically and through a pricing field experiment, that
demand. The results provide two major insights. First, co-branding with the adopting a hybrid automation approach for B2B retail harnesses benefits of sales
closed museum lifts demand for the partner museum. Second, the increased people expertise as well as benefits of automation, depending on the
demand is the product of two counter-acting forces: first, customers with no characteristics of the transaction and the client, and leads to significant increase in
history of buying membership from either museum enter the market, consistent profits. By automating the decision of when to use the expert salesperson and
with the brand providing a quality signal, and second, a sub-group of customers when to use automatic pricing, issues of algorithm aversion and misuse could be
who previously purchased from either or both of the museums display decreased avoided, leading to further increase in profits to the company.
demand consistent with brand dilution. These results have implications for the
treatment of brand intercepts in counterfactuals for co-branding and merger
analysis.
18
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 19
General Session
Chair: Eric Schwartz, University of Michigan, Ann Arbor, MI, 48109-
1234, United States, [email protected]
1 - Bandit Adaptive MaxDiff: Cost-Effective Scalable Idea Screening
Eric Schwartz, University of Michigan, Ann Arbor, MI, 48109-
1234, United States, [email protected], Alexander Zaitzeff,
Kenneth Fairchild, Bryan Orme
Marketing research often measures consumers’ preferences in order to identify
their most preferred items among a set of candidates, such as product benefits,
innovative features, or message appeals. Common techniques are idea screening,
including conjoint analysis and maximum-difference (MaxDiff) scaling. These
methods, however, are not built for large-scale problems. For large MaxDiff
studies whose main purpose is identifying the top few items for the respondent
19
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 20
n TC02
product categories. Review content information has a higher impact on sales
when the average rating is higher and the variance of ratings is lower. Consumers
depend more on review content when the market is more competitive, immature,
Room 33, Alter Hall or when brand information is not easily accessible. A counterfactual simulation
Digital Economy III: Reviews and Reputation suggests that re-ordering reviews based on content can have the same effect as a
1.6% price cut for boosting conversion.
General Session
4 - Consumer Reviews and Regulation: Evidence from
Chair: Georgios Zervas, Boston University, Boston, MA, United States, NYC Restaurants
[email protected] Georgios Zervas, Boston University School of Management,
1 - Does Occupational Licensing Matter in an Online World? Boston, MA, United States, [email protected], Chiara Farronato
Andrey Fradkin, Massachusetts Institute of Technology,
We investigate complementarities between two signals of restaurant quality:
179 Prospect Street, Cambridge, MA, 02139, United States, health inspections and online reviews. To protect consumers, health inspectors
[email protected], Chiara Farronato, Erik Brynjolfsson periodically evaluate restaurant hygiene, and assign health grades to restaurants.
This paper examines how consumers’ choices of service professionals are related Consumers are also able to rate restaurant quality online, through review
to the occupational licensing status of service providers. The setting is a large, US platforms like Yelp. We investigate whether consumer reviews can predict
online labor market platform, where professional service providers bid on hygiene violations. To do so, we implement we use machine learning to predict
consumers’ projects. We outline an econometric approach to estimate the causal individual restaurant violations from the text of Yelp reviews. We find that
effect of a professional’s licensing status and reputation on the consumer’s choice consumer reviews are good predictors of food handling violations, but are poor
of whom to hire. Our findings suggest that a professional’s digital reputation predictors of facilities and maintenance violations. We then investigate how the
matters more than the professional’s licensing status for customer choice. We also hygienic information contained in online reviews affects demand and supply. On
study the market wide effects of occupational licensing stringency on competition the demand side, we find that conditional on hygiene related information
and service quality. contained in online reviews, consumers still take health grades into account when
choosing restaurants. On the supply side, we find that review platforms create
2 - Revealed Preference Heterogeneity Through Online Ratings reputational incentives for restaurants to reduce hygiene violations that are
Brett Hollenbeck, University of California-Los Angeles Anderson, predictable by Yelp reviews. Specifically, we find that relative to restaurants not
1317 S Westgate Avenue, Apt 204, Los Angeles, CA, 90025, on Yelp, restaurants on Yelp score better on hygiene dimensions detectable by
United States, [email protected], Yufeng Huang consumers than on dimensions not detectable by consumers. Our results have
implications for the design of government regulation in a world where consumers
Consumers often rely on product reviews to make purchase decisions, but how rate their service experiences online.
consumers use review content in their decision making has remained a black box.
In the past, extracting information from product reviews has been a labor
intensive process that has restricted studies on this topic to single product
categories or those limited to summary statistics such as volume, valence, and n TC03
ratings. This paper uses deep learning natural language processing techniques to
overcome the limitations of manual information extraction and shed light into Room 34, Alter Hall
the black box of how consumers use review content. With the help of a
comprehensive dataset that tracks individual-level review reading, search, as well When Fashion Meets Marketing Science
as purchase behaviors on an e-commerce portal, we extract six quality and price General Session
content dimensions from over 500,000 reviews, covering nearly 600 product
categories. The scale, scope, and precision of such a study would have been Chair: Yu-San Lin, Penn State University, University Park, PA, 16802,
impractical using human coders or classical machine learning models. We achieve United States, [email protected]
two objectives. First, we describe consumers’ review content reading behaviors.
Co-Chair: Tavy Ronen, Rutgers University, Bridgeton, NJ, United States,
We find that although consumers do not read review content all the time, they do
rely on review content for products that are expensive or of uncertain quality. [email protected]
Second, we quantify the causal impact of content information of read reviews on Co-Chair: Tawei (David) Wang, DePaul University, Chicago, IL, 60604,
sales. We use a regression discontinuity in time design and leverage the variation United States, [email protected]
in the review content seen by consumers due to newly added reviews. To extract
content information, we develop two deep learning models: a full deep learning 1 - The Role of Imagery in Equity Crowdfunding
model that predicts conversion directly and a partial deep learning model that Yusan Lin, Penn State University, State College, PA, United States,
identifies content dimensions. Across both models, we find that aesthetics and [email protected], Miwako Nitani, Tavy Ronen, Tawei Wang
price content in the reviews significantly affect conversion across almost all
This paper examines the role of imagery in crowdfunding from 2014-2016. While
product categories. Review content information has a higher impact on sales
previous papers have documented the correlation between entrepreneurs’ social
when the average rating is higher and the variance of ratings is lower. Consumers
networks and the probability of campaign success, the underlying mechanism has
depend more on review content when the market is more competitive, immature,
not been fully investigated. This paper addresses this gap in the literature by
or when brand information is not easily accessible. A counterfactual simulation
exploring the role of images originating from social media networks in
suggests that re-ordering reviews based on content can have the same effect as a
crowdfunding campaigns. We focus on fashion and beauty related companies and
1.6% price cut for boosting conversion.
argue that the probability of campaign success depending on reliable imagery. By
3 - Large-Scale Cross-Category Analysis of Consumer Review leveraging deep learning neural networks to understand the contents of imagery,
Content on Sales Conversion Leveraging Deep Learning we show that by transmitting reliable imagery regarding the main product line,
Xiao Liu, New York University, Stern School of Business, 44 W 4th fundraising campaigns on crowdfunding platforms are more likely to be
successful. Finally, we document how crowd investors interpret those signals and
Street, New York, NY, 10013, United States, [email protected], use for their investment decisions.
Dokyun Lee, Kannan Srinivasan
2 - Social Competition of Fashion Brands on Instagram
Consumers often rely on product reviews to make purchase decisions, but how
consumers use review content in their decision making has remained a black box. Tawei (David) Wang, DePaul University, Chicago, IL, United States,
In the past, extracting information from product reviews has been a labor [email protected], Yusan Lin
intensive process that has restricted studies on this topic to single product Since Instagram started its service about 7 years ago, it has been a popular social
categories or those limited to summary statistics such as volume, valence, and media outlet for all kinds of fashion brands. For instance, Nike, Victoria’s Secret,
ratings. This paper uses deep learning natural language processing techniques to H&M, and Louis Vuitton all have more than 20 million followers on Instagram in
overcome the limitations of manual information extraction and shed light into 2017. Another example, in the month before the London Fashion Week in 2016,
the black box of how consumers use review content. With the help of a there were more than 5,500 posts with #LFW2016 on Instagram compared to
comprehensive dataset that tracks individual-level review reading, search, as well only about 1,100 Twitter mentions in the same period. These pieces of anecdotal
as purchase behaviors on an e-commerce portal, we extract six quality and price evidence suggest that Instagram has become a crucial part of fashion brands’
content dimensions from over 500,000 reviews, covering nearly 600 product competing fields. In this study, we explore (1) the information posted by fashion
categories. The scale, scope, and precision of such a study would have been brands and (2) the competitive actions (both action timing and types of actions)
impractical using human coders or classical machine learning models. We achieve taken by fashion brands in the context of Instagram. We collected Instagram posts
two objectives. First, we describe consumers’ review content reading behaviors. by 480 fashion companies across a two-year period, and find that the more
We find that although consumers do not read review content all the time, they do consumers two fashion companies share, the more frequently they take actions to
rely on review content for products that are expensive or of uncertain quality. respond each other’s signals. Our preliminary findings would sketch the
Second, we quantify the causal impact of content information of read reviews on competitive landscape on Instagram for fashion brands and provide managerial
sales. We use a regression discontinuity in time design and leverage the variation implications when forming social media strategies.
in the review content seen by consumers due to newly added reviews. To extract
content information, we develop two deep learning models: a full deep learning
model that predicts conversion directly and a partial deep learning model that
identifies content dimensions. Across both models, we find that aesthetics and
price content in the reviews significantly affect conversion across almost all
20
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 21
21
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 22
n TC05
currently employed by Automotive Lubricant companies in India. This research is
exploratory in nature and its scope is to find out the existing process of devising
an integrated marketing communication channel strategy currently in practice by
Room 232, Alter Hall Indian Automotive Lubricant companies. The question broadens its scope to
Panel Discussion on Digital Marketing Applications of understand the selection of the available mediums currently employed by
AI and Deep Learning III Automotive Lubricant Companies.Qualitative Research Method involving Case
Study Method have been used to address the subject. The question broadens its
Invited Session scope to understand the selection of the available mediums currently employed
1 - Panel Discussion on Digital Marketing Applications of AI and by Automotive Lubricant Companies. The two major steps for the data analysis in
this research include Within Case Analysis and Multi Case Analysis.The data
Deep Learning III analysis using grounded theory has been done involving several iterations
Moderator: Xueming Luo, Temple University, 1801 Liacouras between interview data, existing theory and observation data.In the multi case
Walk, Philadelphia, PA, 19122, United States, [email protected] analysis, replication logic is used. This analysis is done to seek comparison and
AI and DL algorithms will leverage big and rich data like sentiment, image, video, explain common, different and complimenting findings in the study of channel
and diverse subtleties of human interactions in one end-to-end process, not selection mechanism of marketing communication by automotive lubricant
captured by traditional marketing. Empowered by deep neural networks, new companies in India. The findings across cases are integrated to form common
marketing applications in the industry are astonishing (academic works are analysis background. Five relevant stages have been found from the empirical
lacking): smart chatbots with personality, consumer classification and unmet data in this research study which includes the new dimensions under which the
needs detection, user preference discovery and recommendation, social semantics processes were carried out.
and personalized content curation, digital fraud and data breaches prevention, 3 - Intra-firm and Inter-firm Spillover Effects in Display
intelligent customer service and ad targeting, omnichannel marketing and
Advertising:implications for Multi-brand Portfolio Firms
dynamic pricing, and autonomous retailing and platform ecosystems.
Lin Boldt, Clark University, 950 Main Street, Worcester, MA,
Panelists 01610, United States, [email protected], Neeraj Arora
Oded Netzer, Columbia Business School, New York, NY, 1027, Advertising spillover effect refers to the phenomenon that the advertising of one
United States, [email protected] brand could not only benefit brand being advertised, but also increase demand for
Arun Sundararajan, New York University, Stern School of Business other brands. In this paper, we examine the spillover effect of online display
Info Sys Dept, 44 West 4th Street KMC 8-93, New York, NY, advertising for firms that own a portfolio of brands in a category. First, who
10012, United States, [email protected] benefits from online display advertising? The advertising brand? The other brands
in the firm’s portfolio? Or the brands from a competing firm? Second, how should
Avi Goldfarb, University of Toronto, Rotman School of a multi-brand firm accurately attribute the effect of each brand’s advertising and
Management, 105 St George Street, Toronto, ON, M5S 3E6, optimally allocate the advertising budget across a portfolio of brands? Third, what
Canada, [email protected] are the moderators of spillover effects? In particular, we focus on intra-firm and
Kannan Srinivasan, Carnegie Mellon University, Pittsburgh, PA, inter-firm product feature and price similarity. We address the questions using a
15213, United States, [email protected] unique household-level panel data, which include each household’s exposure to
randomized display ads from multiple advertisers on media websites and the
Daria Dzyabura, New York University, 44 West 4th Street, household’s browsing behavior on e-commerce websites. Our findings contribute
Tisch Hall, New York, NY, 10012, United States, to the literature in the following ways. First, we provide evidence of the display
[email protected] advertising effect not only on the advertised brand, but also on the other brands
in the firm’s portfolio and competing brands offered by other firms. Second, we
propose an attribution and allocation strategy for multi-brand companies in the
n TC06 face of the intra-firm and inter-firm advertising spillover effect. Third, this paper is
the first in the literature to understand the role of feature similarity and price
Room 234, Alter Hall similarity on the advertising spillover effect for a multi-brand firm. Our proposed
framework provides guidance for multi-brand companies on how best to spread
Advertising Strategies its advertising dollars across the brands in its portfolio.
Contributed Session 4 - The Debate over Loyalty Program Performance: Identifying
Chair: Sreyaa Guha, IE Business School, IE University, Calle Maria De Factors that Drive Success of Coalition Loyalty Programs
Molina 12, Madrid, 28004, Spain, [email protected] Sreyaa Guha, PhD Candidate, IE Business School, IE University,
Calle Maria De Molina 12, Madrid, 28006, Spain,
1 - Using Advertising Disclosure Choices to Predict Firm [email protected], Shameek Sinha
Demand Evolution
Simone Wies, Goethe University Frankfurt, Theodor-W-Adorno Loyalty programs are valuable relationship building tools, especially since the
Platz 3, Frankfurt, 60323, Germany, [email protected] associated customer datasets can be used for targeted marketing. However, the
performance of standalone loyalty programs is debatable with issues of self-
Firms have considerable leeway in deciding whether they report advertising selection, competition and diverse customer requirements. Coalition loyalty
expenses or not. We argue that the reporting of advertising expenses can reveal a programs with extensive category partnerships, variety of rewards and enhanced
firm’s expected demand evolution. Specifically, we theorize that firms might stop customer experience provide a viable alternative. In this research we analyze
disclosing advertising spending when they seek to disguise reduced demand for its strategic dimensions of customer involvement in such programs and identify
products. That is, if the firm sees its demand eroding, it might boost advertising tactical drivers, which if implemented, facilitates program success. The three
spending in an effort to buffer demand shortcomings. We also examine whether strategic behavioral metrics we analyzed were transactions with partners across
investors understand these motivations and interpret ceased advertising disclosure categories, responses to promotional campaigns and redemption of rewards. With
as a firm’s signal of despair to secure sales. To test our predictions, we estimate access to an unique customer panel from a coalition loyalty program, we used
forecasting models for size-adjusted advertising series to identify the difference panel estimation techniques on euros spend on transactions and points used for
between actual and expected advertising levels and to determine whether the redemptions (continuous) and choice of store format in transactions, choice of
firm reports above- or below-normal advertising. We then link the firm’s channels for reward redemptions, choice of response mode in campaign responses
reporting choice and advertising deviations to actual and future performance. and likelihood of campaign impact (discrete) to investigate the tactical drivers that
Using both an event study and a calendar time portfolio approach, we finally test increase customer involvement. We find that not only managing the point offers
to which extent the stock market understands and assesses the signal of demand on transactions and reward offers for redemptions are critical, but also
evolution. Our study contributes to the ongoing discourse of whether and how to understanding the customer profile plays an important role. Promotional
report marketing investments, and outlines how advertising reporting can serve strategies and their timing need consideration. Firms should analyze the partners
as (unintentional) signal of a firm’s demand evolution. and channels where customers transact, respond and redeem. The managerial
insights from this research will help firms efficiently retain and acquire customers
2 - Integrated Marketing Channel Selection Mechanism of under coalition loyalty programs. This research also provides theoretical support
Automotive Lubricant Industry in India- An Exploratory Study for investigating all three behavioral metrics simultaneously.
Meenakshi Tomar, O P Jindal Global University, Sonepat Narela
Highway, Village:Jagdishpur, Sonepat, 131001, India,
[email protected]
In Indian context the Automotive Lubricant Marketers have been relying on
traditional mediums for promotional aspects, creating awareness, building brands
and approaching new markets. They now need to study the modern mechanism
used by people in their daily lives to communicate with each other. The
Automotive Lubricant companies in India continue to witness a static growth
which has proved to be a hurdle to increase their market share because of their
reliability on traditional mediums. Therefore an attempt has been made to arrive
at the channel selection mechanism of Integrated Marketing Communications
22
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 23
n TC07
that specializes in second-hand property trading. An agent network is formed
around a specific property that includes multiple agents of different roles on the
homeowner side and multiple agents of different roles on the homebuyer side.
Room 237, Alter Hall The structure of the agent network is dynamic, constantly changing with the
Sales & Salesforce Management I arrival of a new potential buyer and her associated agents. To overcome the
endogeneity concern that is typical of agent network formation, we take
Contributed Session advantage of the real estate platform’s special policy and the exogenous arrival of
Chair: Junhong Chu, National University of Singapore, Nus Business property buyers. Lianjia has been practicing the “gong-fang-si-ke” (“public-
School, 15 Kent Ridge Drive, Singapore, 119245, Singapore, property-private-customer”) policy that stipulates (1) all properties, regardless
how they are acquired and who acquires them, are the company’s public
[email protected] resources (or “commons”) and managed by the specific trading zone’s storefront
1 - Free Riding Effect Between Sales People in Adoption of a and can be accessed by all buyers’ agents, and (2) a property buyer is the private
Complex New Product customer of the agent who acquires the buyer. Both the arrival of a property
Vahideh Sadat Abedi, Cal State Fullerton, 3099 W Chapman Ave, buyer and her preferred location for a property are exogenous to Lianjia and its
Apt 259, Orange, CA, 92868, United States, [email protected], agents, which help generate exogenous agent networks. Further, the same agent
Rahul Bhaskar can play different roles in different properties acting on both the buyers and the
homeowners. These special data structure and institutional arrangements enable
Purchase of complex new products (e.g. insurance plans) typically depends on us to separately measure the effects of agent network structure, different agent
both customer word-of-mouth about the product and about the service offering roles, and each agent on the outcomes of a property, and identify the optimal
brokers. We show that this leads to a free-riding effect between the brokers. We agent network size to maximize the sales outcome.
model this sales process, and validate it from data for an insurance product. We
measure the free-riding effect, and show how the resulting model can facilitate
decision making in managing the sales force.
2 - Effect of Sales Momentum on Salesperson Performance
n TC08
Irene Nahm, Doctoral Student, University of Houston, Room 238, Alter Hall
4750 Calhoun Road, Houston, TX, 77204, United States, CB - Time, Tourism & Culture
[email protected], Michael Ahearne, Seshadri N Tirunillai, Nick Lee
Contributed Session
Inside sales force are agents that mostly reach customers remotely, usually aided
by technology. This research examines the role of momentum in managing Chair: Ozum Zor, Rutgers Business School, Piscataway Township, NJ,
individual sales performance for inside sales force. We argue that salesperson 0885, United States, [email protected]
experience periods of increased (or decreased) performance, when compared with
1 - Time Use and Purchase Behavior
their expected performance. We refer to this short-term persistence in a
salesperson’s performance as (sales) momentum. We examine momentum using a Yan Xu, Assistant Professor, The Hong Kong Polytechnic
novel, high-frequency transactional dataset from a large European call-center University, Hong Kong, China, [email protected],
firm. The results provide a strong evidence for the existence of momentum and Bart Bronnenberg, Tobias Klein
indicate that momentum is not an inherent trait of a salesperson that makes a Consuming goods not only costs money to buy market goods, but also time spent
specific salesperson prone to clumps of successful (or failed) outcomes. Rather, on home production broadly defined, i.e., on shopping, evaluating products,
momentum is a state some salespeople transition into under appropriate preparing meals, and consumption. Consequently, time availability affects how a
conditions. We also examine how momentum influence individual salesperson’s household uses the market and which bundle of market goods it chooses. Using a
performance over time and how social factors moderate this relationship. We unique individual level data set combining purchase records, household
exploit the variation across the two call-center locations to understand the characteristics, and home production preferences, this paper studies how
importance of social factors in sustaining the momentum in sales performance plausibly exogenous shocks on time-availability impact purchasing behavior. We
and find that accounting for both social effects and momentum is important in classify a subset of products according to the time it takes to turn them into
explaining salesperson performance. The results suggest an asymmetry in the consumption experiences. Controlling for household fixed effects, we show that
influence of social factors on momentum. While social factor has a strong increases in the availability of time from, e.g., retirement lead to more shopping
influence in negative momentum, it does not help sustain positive momentum. trips, a more diverse set of stores, more overall spending on groceries, and more
These findings suggest that managers can enhance salesperson performance by variety purchased. In line with our main hypothesis, time availability also shifts a
detecting momentum. Further, managers could significantly alter the win rates by household’s shopping bundle into more time-intensive goods. Therefore, it is
using timely intervention to break the negative momentum of a salesperson. important to account for the availability of time when analyzing consumer
3 - A Holistic View of Sales Effectiveness an Empirical Investigation purchase behaviors. Firms can make their products attractive to the consumer not
Srinath Gopalakrishna, Professor of Marketing, University of only by lowering the price but also by making products more convenient to use.
Missouri, Trulaske Sr. College of Business, 434 Cornell Hall, 2 - The Impact of Inbound Tourist Lifestyles on Travel Behaviors
Columbia, MO, 65211, United States, [email protected], Yusuf Erkaya, International Burch University, Sarajevo,
Ashutosh R Patil, Andrew Crecelius Bosnia and Herzegovina, [email protected],
Researchers and practitioners agree that customer acquisition and retention are Teoman Duman, Omer Topaloglu
critical aspects of sales success. Salespeople spend considerable time across these Despite the fact that understanding tourist motivations and lifestyles is important
two activities which are respectively known as “hunting” and “farming” in the in devising marketing strategies for tourism service providers, research in these
sales literature. There is general agreement that the effectiveness with which each areas with respect to inbound tourism is very limited. Overwhelming majority of
activity is performed varies considerably across salespeople. Some are better at recent research studies on tourist motivation, lifestyles, and behavior have
acquisition, others are better at retention, while still others are effective at both focused on demand for mass tourism vacations. Relying on the lifestyle theory,
activities. We develop a composite sales effectiveness metric and undertake a this paper focuses on inbound tourism where local and regional tourism activities
systematic empirical investigation of the drivers of each component as well as the have been analyzed to formulate marketing strategies for tourism service
drivers of overall sales effectiveness. We find that such an investigation that providers based on an investigation of inbound tourists’ motivations and
examines the comprehensive effectiveness at both these activities is missing in lifestyles. Data for this study include 575 self-administered surveys completed by
extant literature. Based on reviewing relevant theories, we develop a bivariate inbound tourists in Vienna, Austria. Austrian inbound tourism provides a typical
model of hunting and farming effectiveness. We propose and test several context that is present in few European destinations. As an inbound tourism
hypotheses involving the drivers - notably we consider impersonal destination, Austria serves as an ideal setting because of its location and its ability
communication, operationalized by advertising spending and personal contact, to satisfy various types of touristic motivations, such as historic, family, summer
operationalized by number of customer service personnel. We assess the impact of vacation, etc. The data is analyzed by structural equation modeling. The results
these two variables, among others, and highlight the trade-off that occurs reveal twelve main types of tourists in the inbound tourism sector. Local food and
between these two activities as salespeople balance acquisition and retention local life-motivated tourists emerged as the most prevalent tourist types. Besides,
efforts. We offer several managerial implications from our research that can help relaxation-pleasure seeking, romance seeking, novelty-adventure seeking, and
managers understand the impact of salesperson, territory, and marketing variables peer acceptance seeking are the other four most dominant types of tourists. In
on hunting, farming and overall sales effectiveness and offers some insights into addition, this study shows that the motivational lifestyle characteristics depend on
how they can deploy resources in an effective manner. travel frequency, household income, time spent for travel, the latest vacation,
4 - Agent Network Structure and Performance in the Second-hand average vacation period, accommodation preferences as well as demographic
characteristics of inbound tourists. Managerial and theoretical implications of the
Property Market findings are discussed.
Junhong Chu, National University of Singapore, Nus Business
School, 15 Kent Ridge Drive, Singapore, 119245, Singapore,
[email protected], Xianling Yang, Li Wang
We study how agent network structure affects the outcomes of second-hand
properties, including the probability to sell, time on market, and sale price, using
proprietary data from Lianjia, one of China’s largest real estate platform agencies
23
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 24
24
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 25
n TC10 n TC11
Room 605, Alter Hall Room 606, Alter Hall
Corporate Social Responsibility I Digital Marketing III
Contributed Session Contributed Session
Chair: Chandra Srivastava, University of Texas-Austin, 4711 Avenue G, Chair: Jungju Yu, Yale University SOM, 186 Edwards St, Apt 2r,
Austin, TX, 78751, United States, [email protected] New Haven, CT, 06511-3780, United States, [email protected]
1 - Forced to do Good: Economic Consequences of CSR Mandates 1 - On the Choice Between Open and Private Ad Exchanges in Online
Nandini Ramani, University of Texas-Austin, 800 West 38th Street, Display Advertising
Unit 4206, Austin, TX, 78705, United States, Amin Z Derakhshi, University of Washington, 840 E Denny Way
[email protected], Raji Srinivasan Apt 206, Seattle, WA, 98122, United States, [email protected],
While firms’ corporate social responsibility (CSR) initiatives are, for the most part, Amin Sayedi
voluntary and government intervention in CSR initiatives may seem antithetical, Display advertising impressions in real-time bidding are bought and sold in two
in an attempt to achieve inclusive development, some governments (e.g., China, types of marketplaces: open ad exchanges and private ad exchanges. An open ad
India, and the United Kingdom) are enacting policy to encourage firms to engage exchange is accessible to all advertisers and publishers who want to buy and sell
in CSR initiatives and even, going so far as to mandate firm CSR spending. A impressions. However, only a prescreened set of advertisers and publishers have
priori, it is not clear what the effects of such government intervention in CSR access to a private ad exchange. We investigate the pros and cons of the two ad
spending are on firm behavior and performance. Does government intervention exchange types and realize that open exchanges pose a bilateral adverse selection
in CSR spending alter a firm’s strategy? How can firms make strategic use of CSR problem, whereas private exchanges soften the competition among advertisers.
mandates? In this paper, we examine how firms can use marketing to strategically We identify the heterogeneity among advertisers and the heterogeneity among
improve their performance in the light of mandated CSR spending. Combining publishers as the two key market factors that could impact the choice of an ad
developments in stakeholder theory and agency theory, we hypothesize exchange type in equilibrium. We build a game-theoretic model to investigate the
heterogeneous effects of mandated CSR spending on firm performance based on strategies of vertically differentiated advertisers and publishers in the market. We
firms’ consumer orientation and ownership structure for two groups of firms - find that, in equilibrium, publishers set a relatively high reserve price in the
those who were already spending on CSR before the mandate, and those who are private exchange and sell an impression to the winner of an open exchange only
spending on CSR for the first time after the mandate. We exploit a 2013 CSR if the impression is left unsold in the private exchange. We also find that when
mandate in India which required some firms to spend two percent of profits on the heterogeneity among the publishers is high and that among the advertisers is
CSR to identify the effects of mandated CSR spending on performance. Our low, all advertisers bid in the open exchange. Conversely, when the heterogeneity
findings indicate that increasing CSR spending following the mandate improves among the publishers is low and that among the advertisers is high, different
performance for firms’ with high spending on advertising, R&D, and distribution. advertisers bid in different types of ad exchanges. Implications for the auditing
Further, increasing CSR spending following the mandate improves performance and monitoring practices in the display ad industry are discussed.
for firms belonging to business groups. Our findings suggest that marketing can
help firms strategically use mandated CSR spending to improve performance, and 2 - Can Bad-quality Images Lead to Greater Demand on Airbnb?
provide guidance to firms on optimal CSR and marketing spending levels to Nitin Mehta, University of Toronto, 105 St George Street, Room
maximize shareholder value. 5072, Rotman School of Management, Toronto, ON, M5S3E6,
Canada, [email protected], Shunyuan Zhang,
2 - Corporate Social Responsibility and Product Quality:
Param Vir Singh, Kannan Srinivasan
The Role of Social Comparison Effect
Moein Javadian, Rotman School of Management, This study answers the question of ‘Why are Airbnb hosts using low-quality
University of Toronto, Toronto, ON, Canada, property images?”. Existing studies on Airbnb show that high-quality images,
[email protected] compared to low-quality images, generate more demand for the Airbnb
properties. However, a large number of Airbnb hosts use low-quality images—
The genesis of CSR as an instrument to attract consumers and increase market even when Airbnb offered a professional photography program for free. This is
power has raised a question on how it interacts with conventional components of because high-quality images come with a risk of creating a high expectation on
marketing strategy. In particular, the relationship of product or service quality is the delivered lodging experience for the guests. As a result, high-quality images
critical because brand managers need to allocate limited resources across these increase the ‘gap’ between the images-induce expectation and the realization, and
two levers to optimize profitability. In a nutshell, it is critical for the manager to hence reduce the likelihood that a consumer leaves a review. High-quality
understand the relationship between CSR and quality in order to deliver the images, though boost demand in the short-run, may hurt hosts in the long-run, if
correct combination to the marketplace. CSR is not simply another dimension of the hosts cannot deliver a lodging experience that meets the expectation and
quality, but they differ in the sense that CSR can have a strong impact on a grow the reviews. Hence, a host may not want to have (free) high-quality images,
person’s image in social context. The extent to which social comparisons influence if his marginal cost of providing high-quality service (e.g., respond to guest
preferences for CSR is then controlled by consumption visibility. In this context, I promptly) is high. We build a structural model on both the demand and the
look at the impact of consumption visibility on the relation between CSR and supply side of the platform. On the demand side, leveraging GPU parallel
quality. More specifically, the question is whether quality and CSR are substitutes computing, we estimate a random-coefficient logit aggregated-demand model,
or complements when the strength of social comparison effect due to with heterogeneous consumers choose from over 2,000 differentiated lodging
consumption visibility is low or high. alternatives. On the supply side, we build a dynamic game model, with
heterogeneous Airbnb hosts dynamically choose the optimal quality of images
3 - Balancing Act: Effect of Female Power in the Top Management
and effort of service to maximize their expected long-term profits. Our analysis
Team on Investments in Marketing reveals interesting (reviews-induced) separating equilibrium, where hosts who
Chandra Srivastava, University of Texas at Austin, 4527 North are able (unable) to provide good lodging experience choose high-quality images
Lamar Blvd., Apt. 5139, Austin, TX, 78751, United States, and serve ‘high-type’ (‘low-type’) consumers. Counterfactual analysis
[email protected], Saim Kashmiri, Vijay Mahajan recommends an image program for Airbnb to improve the revenues. The
proposed policy outperforms Airbnb’s current photography program, by providing
A substantial body of work has illuminated the effect that marketing has on firm
images with medium quality to hosts who would use low-quality images under
performance. However, less research has investigated marketing’s antecedents
the current policy. The medium-quality images generate more demand for the
and what factors make firms more likely to invest in marketing in the first place.
hosts, without creating a big dissatisfaction.
The authors build on upper echelons theory and agency-communion theory to
hypothesize that gender balance in the top management team (TMT) positively
influences a firm’s attention to customer relationships and its investments in the
marketing function. A longitudinal analysis of 221 publicly traded U.S. Fortune
500 firms from 2007 to 2013 supports the authors’ hypotheses that a relative
increase in female power in the TMT is positively associated with CMO presence,
advertising intensity, and corporate social responsibility (CSR), and that
advertising intensity and CSR, in turn, mediate the positive relationship between
female power in the TMT and firm performance.
25
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 26
26
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 27
n TC13
level purchase records in physical stores. Using a doubly robust estimator that
incorporates nonparametric machine learning methods, we find that receiving
email ads can increase a consumer’s spending in physical stores by approximately
Room 746, Alter Hall $11.82. Additionally, we find that the increased offline sales result from increased
Multi-Channel Marketing I purchase probability and a wider variety of products being purchased. Further, we
demonstrate that the effect of email ads is heterogeneous across different
Contributed Session consumer segments using a data-driven approach. Interestingly, the effect is
Chair: Vibhanshu Abhishek, Carnegie Mellon University, 4800 Forbes highest among consumers who have fewer interactions with the focal retailer
Avenue, 3024 Hbh, Pittsburgh, PA, 19104, United States, [email protected] recently (i.e., lower email opening frequency). Overall, our results suggest a
reminder effect of email ads. Receiving email ads from the retailer can generate
1 - For Better or for Worse: The Halo Effects of Online Marketplaces awareness and remind the consumer of the retailer’s offerings of various products
on Entrenched Brick-and-Mortar Stores and services, which gradually increase the consumer’s purchase probability in the
Zhiling Bei, University of North Carolina at Chapel Hill, McColl retailer’s physical stores. These findings have direct implications for marketers to
Building Suite 5103, 300 Kenan Dr, Chapel Hill, NC, 27599, improve their digital marketing strategy design and for policy makers who are
United States, [email protected], Katrijn Gielens interested in evaluating the economic impact of prevalent email advertising.
27
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 28
28
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 29
29
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 30
n TD03
person reading a given article, how textual features of a given paragraph (i.e., the
words used) shape whether someone continues to the next paragraph or not.
Results indicate that how much emotion content evokes, whether it uses familiar
Room 34, Alter Hall or vivid language, textual complexity, and whether the words indicate cognitive
Blockchain Technologies and their Applications processes are occurring are all linked to longer reads. Importantly, however, not
to Marketing all emotion increases reading. Consistent with research on appraisal tendencies
and attention, while text that evokes anger and sadness discourage further
General Session reading, text that evokes anxiety encourages it. To address selection concerns, we
control for various features such as what platform the user is reading the article
Chair: Catherine Tucker, Massachusetts Institute of Technology, , (e.g., mobile vs. desktop), the topic of the article, and even article level textual
Cambridge, MA, 02142, United States, [email protected] features. The results shed light on the psychology of content consumption and
Co-Chair: Christian Catalini, Massachusetts Institute of Technology, has implications for content marketing and content producers.
Cambridge, MA, 02142, United States, [email protected] 2 - Effects of Product Innovativeness and Engagement with Online
1 - Blockchain Technology and Marketing: Promise and Pitfalls Users on Branding Strategies: Evidence From Twitter
Christian Catalini, Massachusetts Institute of Technology, Meltem Kiygi Calli, Assistant Professor, Kadir Has University, Cibali
Cambridge, MA, United States, [email protected], Catherine Tucker Campus, Faculty of Management, Fatih, Istanbul, 34083, Turkey,
[email protected], Abdullah Onden
This paper is an overview of the variety of ways that blockchain technologies may
be used to revolutionize marketing. After providing an issue introduction to what Mining social tags enables marketing researches to understand brand associations,
blockchain technologies actually are and what makes them different, we move to competitive market structure and to predict firm performance. Many forms of
their application in marketing. We highlight in particular their potential to online content such as web links, images, photos, videos, and tweets can be
provide costless decentralized verification of digital identities, and the variety of tagged and used for marketing research. This study presents an approach to
applications that this might have to the digital advertising industry. We also gather and analyze social tagging data and shows how marketers can derive
introduce the Cryptoeconomics Initiative at MIT. useful outcomes from these data. In this paper, we investigate the impact of high-
technology (hereafter HT) companies’ engagement with the online users on
2 - Seeding the S-Curve? The Role of Early Adopters in Diffusion brand perception. Firstly, we analyze the overall sentiment trends for each HT
Catherine Tucker, MIT, Cambridge, MA, 02142, United States, brand using collected tweets posted on Twitter which is one of the popular social
[email protected], Christian Catalini media platforms today. Secondly, we analyze the impacts of engagement between
In October 2014, all 4,494 undergraduates at the Massachusetts Institute of companies and users on sentiments related with the brand image and perception.
Technology were given access to Bitcoin, a decentralized digital currency. As a In the study, we collect the Twitter data on five HT product categories. Sentiment
unique feature of the experiment, students who would generally adopt first were analysis is conducted using Sentiment140. It measures the strength of sentiments
placed in a situation where many of their peers received access to the technology by giving scores. To analyze the relationship between engagement and sentiment
before them, and they then had to decide whether to continue to invest in this on social media, we collect the tweets posted by the companies’ or their
digital currency or exit. Our results suggest that when natural early adopters are support/help services’ and analyze them. The impacts of level, media types (link,
delayed relative to their peers, they are more likely to reject the technology. We text, video, photo), length, time (day-of-week and time-of-day) of tweets on
present further evidence that this appears to be driven by identity, in that the customers’ emotions and sentiments are investigated. We also examine the word
effect occurs in situations where natural early adopters’ delay relative to others is frequencies and show word mapping for each brand. Moreover, we analyze the
most visible, and in settings where the natural early adopters would have been tweets about the target HT product innovations for each brand. The results show
somewhat unique in their tech-savvy status. We then show not only that natural that on which technical characteristics of the HT products consumers do focus.
early adopters are more likely to reject the technology if they are delayed, but Examples of the technical terms are face ID, all screen, TrueDepth camera, water
that this rejection generates spillovers on adoption by their peers who are not resistance, wireless charging.
natural early adopters. This suggests that small changes in the initial availability of 3 - Advertising on Premium or Non-premium Websites?
a technology have a lasting effect on its potential: Seeding a technology while Effects on Recall, Attitudes, and Click Behavior
ignoring early adopters’ needs for distinctiveness is counterproductive.
Edlira Shehu, University of Southern Denmark, Campusvej 55,
3 - Mobile Fintech: How Instantaneous Analytics Changes Consumer Odense, DK-5230, Denmark, [email protected], Nadia Abou
Decision Making Nabout, Michel Clement
Shuyi Yu, MIT, Cambridge, MA, United States, [email protected] Hundreds of brands (e.g., AT&T and Verizon) stopped advertising on YouTube and
The Financial services industry has been transformed by the advent of Fintech, the Google display network in 2017 due to fear that their ads would appear on
and especially the use of mobile applications to deliver instantaneous analytics to non-premium websites with inappropriate content. Thus, the inherent quality of
consumers. It is not clear, though how this increase in analytics availability affects websites has become an important determinant of how online marketers plan
decision making. To analyze this I use unique detailed data that tracks financial their online advertising campaigns. Managerial practice currently envisions that
trades made on desktops and mobile. I find suggestive evidence that people uses advertising on premium websites is primarily relevant for branding campaigns,
fewer pieces of data when making purchase decisions on the mobile platform. To but less so for performance campaigns. Understanding how premium and non-
identify whether this is causal, I exploit variation in mobile usage related to premium websites affect online advertising effectiveness has substantial
people reaching mobile data caps at the end of the month. managerial relevance; yet, empirical insights are scarce. We use data from two
large-scale field studies, alongside observational data, to analyze how advertising
on premium and non-premium websites influences advertising effectiveness
along the hierarchy-of-effects. We show that website quality is relevant for both
n TD04 branding and performance campaigns, depending on brand image. For high-
image brands, advertising on premium websites affects both branding and
Room 35, Alter Hall performance campaign effectiveness positively. In contrast, low-image brands do
Digital Marketing II not benefit from advertising on premium websites.
Contributed Session
Chair: Edlira Shehu, University of Southern Denmark, Geschwister-
Scholl-Strasse 64, Hamburg, 20251, Germany, [email protected]
1 - What Leads To Longer Reads? Reading Depth In Online Content
Wendy W. Moe, Robert H. Smith School of Business, University of
Maryland, College Park, 3469 Van Munching Hall, College Park,
MD, 20742, United States, [email protected],
Jonah Berger, David A. Schweidel
In today’s digital age, more and more consumers read content online. They scan
articles on the Wall Street Journal’s website, catch up on sports at ESPN.com, and
peruse blogs on tech and celebrity gossip. But what leads to longer reads? That is,
what about certain articles encourage people to keep reading rather than leave
the article and do something else? To address this question, we use natural
language processing to analyze half a million reading sessions from over 30,000
articles from nine major online publishers. This allows us to examine, for a given
31
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 32
n TD05
experiment field data on a platform reputation policy (Food Safety and Liability
Insurance) in a home-cooked food sharing context. The platform saw an overall
increase of sellers’ sales revenue and buyers’ expenditure after implementing the
Room 232, Alter Hall policy. More interestingly, the insurance policy led to greater sales revenue
Mobile, Algorithm, and Artificial Intelligence (AI) increases for sellers with lower or no review ratings than sellers with higher
Session I: Omnichannel Adverstising and ratings. The results indicate that the reputation policy intervention boosted the
revenue growth of zero-rating and lower-rating sellers by 172% and 162%,
Field Experiments respectively, while the revenue growth of higher-rating sellers was 128%. On the
General Session buyer side, the insurance policy induced the novice buyers to spend more on the
platform than seasoned buyers. The policy intervention boosted inexperienced
Chair: Xueming Luo, Temple University, Philadelphia, PA, 19122, buyers’ expenditure by 395%, whereas there was hardly any increase in
United States, [email protected] experienced buyers’ expenditure. These findings suggest that the platform
reputation enhancement policy grows the sharing economy business through the
Co-Chair: Anandasivam Gopal, University of Maryland, College Park, mechanism of engendering more sales of the low-rating sellers and increasing
MD, 20742, United States, [email protected] more purchases from novice buyers.
1 - Micro-Giving: On the use of Mobile Devices and Monetary
Subsidies in Charitable Giving
Anandasivam Gopal, University of Maryland, University of n TD06
Maryland, College Park, MD, 20742, United States,
[email protected], Dongwon Lee, Dokyun Lee Room 234, Alter Hall
Mobile devices are increasingly being used by non-profits and charitable Analysis of Markets I – Structure and Alliance
organizations as an alternative channel for philanthropy. Specifically,
organizations can construct fund-raising campaigns through mobile applications
Contributed Session
or through mobile services such as SMS messages to target individuals and Chair: Sungtae Eun, Texas Tech University, 5402 66th st, 107, Lubbock,
incentivize charitable giving. Research studying how mobile devices may be used TX, 79424, United States, [email protected]
to enhance such campaigns is limited. In this paper, we study how charitable
giving through mobile apps can be enhanced through the use of push 1 - Modelling and Forecasting the Dynamics of Mobile Devices
notifications and economic incentives. Using a randomized field experiment Market Shares
conducted in partnership with a mobile service provider in the US, we examine Ivan Svetunkov, Assistant Professor, Lancaster University,
the effects of push notifications, monetary subsidies, and intertemporal choices of Lancaster University Management School, Lancaster, LA1 5ET,
rewards on donation outcomes by individual users. Even though the donated United Kingdom, [email protected], Victoria Grigorieva,
amounts are small, consistent with cause marketing campaigns, the campaigns Yana Salichova
are sustainable since the aggregate amounts received are significant over time.
Our experiments provide three main findings. First, push notifications have a The convergence of mobile telephony, Internet services, and personal computing
remarkably high effect on donation behavior in terms of the decision to donate as distorts the mobile devices market. This market can nowadays be considered as a
well as donation amounts, especially for idle users. Second, offering rebates as living ecosystem, where sales of one devices influence the sales of the others.
economic incentives are more effective than offering matching grants, in These days consumers tend to base their purchase decision not solely on the
particular for male consumers. Finally, while users are indifferent between brand loyalty towards a manufacturer but also on the familiarity with the
receiving their rewards now versus later, users that are offered a rebate respond software platform. For example, consumers owning Apple iPhones would rather
with greater alacrity when rewarded immediately rather than later. Our findings buy Apple iPad than a tablet of another competing company, because they are
have important implications for charitable organizations and application service familiar with the operating system and can enjoy the continuity. These
providers as well as for the design of cause marketing campaigns using the mobile connections between the devices create new patterns in the dynamics of market
channel. shares and should be taken into account when companies make marketing
decisions about the development of their devices and software. We propose a
2 - Cross-channel Spillovers and Cross-media Synergies in multivariate model that takes this complex dynamics into account. This approach
Omnichannel Advertising Response not only allows us to capture the interdependencies in the market of mobile
Min Tian, University of Wisconsin, 975 University Ave, Madison, devices, but also generates more accurate forecasts, which supports decision
WI, 53706, United States, [email protected], Paul R. Hoban, making. We test our model on the dataset and compare its performance with
Neeraj Arora several other benchmark models, showing the advantages of the proposed
approach.
Effective management in the modern omnichannel environment, where firms
regularly advertise and sell products both online and offline, requires an 2 - The Impact of Economic Crises on Alliance Portfolio Composition
understanding of potential cross-channel spillovers and cross-media synergies. Tuba Yilmaz, Assistant Professor, BI Norwegian, Nydalsveien 37,
Cross-channel spillovers may occur when digital advertising influences offline Oslo, 0484, Norway, [email protected]
shopping, and when traditional media influences online purchase behavior.
An important gap in the alliance literature is the investigation of how business
Cross-media synergies occur when response to advertising in one medium (i.e.,
cycles affect firms’ collaboration behavior. Though it is well known that firms
digital or traditional) is amplified or mitigated because of advertising in the other.
often use alliances to respond to uncertainty and share risks, little is known how
Using a large-scale, randomized field experiment, we explore the effects of
firms adjust alliance strategies in response to economic crises. When faced with
traditional (i.e., direct mail) and digital (i.e., retargeted display) advertising on
rapid, often unforeseen changes in their external environment, firms could need
online consideration and sales, as well as offline purchases. Results indicate
to adapt alliances throughout their alliance portfolio. Alliances can be one of the
significant spillover in online consideration; direct mail increases both the
mechanisms that help firms to reap the opportunities unleashed by the external
incidence and frequency of website visits. With respect to sales, we find significant
change. However, these external shocks often negatively affect the value and the
cross-channel spillover and cross-media synergy among digitally active
effectiveness of alliances and thus makes them more susceptible to failure. This
consumers. Retargeted display advertising increases both online and offline
research explores how firms adapt their alliance portfolio composition following
revenues. Interestingly, when direct mail is served in isolation, it increases offline
an economic crisis. This study then explores how the crisis differentially affect the
revenues, but at the expense of online sales. In contrast, when direct mail is
viability of different forms of alliances.
reinforced by retargeted display advertising, this channel switching behavior
disappears; both online and offline sales increase. Our findings highlight the need
to consider cross-channel spillovers and cross-media synergies to characterize
advertising response accurately.
3 - Platform Protection Intervention Effect in the Sharing Economy
Siliang Tong, Temple University, Temple University, Philadelphia,
PA, 19144, United States, [email protected], Xueming Luo,
Lin Zhijie, Cheng Zhang
Despite the promise of sharing economy, information asymmetry still exists
where sellers possess more knowledge of product quality than buyers do. Thus,
the surge of sharing economy puts enormous pressures on platform reputation
management. While many platforms enact consumer protection policies such as
purchase insurance, empirical evidence regarding the mechanisms on how such
policy would take effect remains limited. This study exploits a rich natural
32
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 33
33
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 34
34
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 35
36
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 37
4 - How do Information Strategy Online Lead to a Win-Win Situation 3 - Image Network and Interest Group a Heterogeneous Network
with Free Riding Problems Embedding Approach to Analyze Social Curation on Pinterest
Yimeng Li, PhD, University of East Anglia, 3 Cunningham Road, Kunpeng Zhang, University of Maryland, 4347 Van Munching
Norwich, NR58HG, United Kingdom, [email protected], Hall, 4316, College park, MD, 20742, United States,
Franco Mariuzzo, Nikolaos Korfiatis [email protected], Liye Ma, Baohong Sun
Recent foreclosures of high street retail shops further challenge the viability of Social curation has become an effective process for consumers to navigate
brick-and-mortar retailers in an era dominated by electronic commerce. Classic through the vast digital content to find product information that fits their
brick-and-mortar shops rely on the reputation of facilitating customer interests. However, little is known about the characteristics and implications of
understanding of a product through expert sales assistance. On the other hand, the social curation process, and there is little guidance on how brands should use
the internet allows sellers to operate online stores at lower cost with a structured social curation tools to achieve marketing goals. Using the popular image curation
provision of information. Online stores can also free-ride on the investment platform Pinterest.com as the empirical context, this research aims at
undertaken by physical stores, e.g. the provision of unit inspections to the understanding (i) how digital content are collected, classified and curated by
customer and personalized assistance, even he provides during his product search. users; (ii) what these activities reveal about the latent preference of users. To do
Studying competition in the information collection stage and purchase stage so, we leverage heterogeneous network embedding, a powerful technique
between the two channels, we find that under certain conditions the occurrence recently developed in machine learning. We treat images, curation keywords, and
of free-riding is ultimately beneficial to brick-and-mortar retailers, not the users as nodes in a heterogeneous network, where edges are defined by users’
expected online sellers. Information gathered through these two stages can also collection and curation actions. The heterogeneous network embedding
lead to cross-channel browse-and-switch behavior. This paper demonstrates that technique captures structural and semantic correlation among multiple types of
this behavior always intensifies competition and does harm to both sellers’ profits, node in large-scale networks. Aligning image and user networks in the same
however in certain conditions the e-tailer can take the strategy of affecting space allows us to relate image classification with user interests and the context
consumers’ switching cost to soften competition between the sellers, minimizing revealed in curations. Applying the approach to a large data set of users and
damage to their own profits. images from Pinterest.com, our results show the images are effectively organized
based on features and user curation. The underlying structure and relationship
revealed from the embedding analysis show that users grouped into same clusters
tend to curate similar images. We further demonstrate how the technique can be
used to assess brand perception and to customize image design to improve user
acceptance.
37
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 38
38
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 39
n TD17
consequences of consumer motivation using experimental research paradigms
and shed light on many theoretical facets of motivation. While some of these
findings can be applied to consumer welfare, this research has not inspired much
Room 31, Alter Hall application by practitioners. This could be a result of the difficulty in uncovering
Panel Session: ISMS Code of Conduct Town Hall motivations and examining their impact for specific firms. The present research
Discussion puts forth a practical technique by which firms can extract consumer motivations
from customer text responses. Specifically, we extract motivations from 25,290
Panel Session hosts who list properties on the Airbnb platform by mining their text responses to
the question, “why did you start hosting?”. We find that Airbnb hosts are driven
Chair: Sandy Jap, Emory University, Atlanta, GA, [email protected] by three principle motivations - the motivation “to earn cash,” “to share beauty,”
1 - ISMS Code of Conduct Town Hall Discussion and “to meet people.” We find that these motivations have downstream
Moderator: Sandy Jap, Emory University, Atlanta, GA, consequences for hosts’ engagement with the platform and likelihood of staying
[email protected] on the platform. We use different analytical tools including machine learning and
natural language processing, “buy-till-you-die” latent attrition model and lab
Discussion about ISMS Code of Conduct experiments. These findings suggest that firms can leverage these motivations in
Panelist: Leigh McAlister, University of Texas, Austin, TX, their own advertising by attracting the right “type” of customer, which could also
[email protected] dictate firm revenue.
Barbara Kahn, University of Pennsylvania, Philadelphia, PA, 3 - Generosity or Sacrifice – Which Signal Derives Higher
[email protected] Monetary Donations
Gene Anderson, Syracuse University, Syracuse, NY, Gil Peleg, University of the Negev, Beersheba, Israel,
[email protected] [email protected], Oded Lowengart, Danny Shapira
Dan Airely, Duke University, Durham, NC, [email protected] Often donors take into account the social recognition component as part of the
decision-making process regarding the level of contribution they make. Since
societies value the economic value of the donation, as well as the benefactors’
sacrifice (or effort), beneficiaries can incentivize donors by publishing their
donations (and effort) as a signal to society. In this study, we analyzed various
signaling procedures that trigger different social reward schemes in order to gain
better fundraising outcomes. We developed a pseudo principal-agent model to
reveal the best strategy for beneficiaries when they use society’s reward as a
motivator for monetary donation. We considered three social reward schemes in
our model: surreptitious donations, generosity signals, and sacrifice signals. The
model yielded counter-intuitive trade-offs as the donor’s wealth become a
moderator for social reward scheme. We conducted a confirmatory lab
experiment to examine the model’s assertions.
39
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 40
40
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 41
41
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 42
42
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 43
43
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 44
44
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 45
Room 606, Alter Hall 4 - Dynamic Structural Modeling of Online Reputation Systems: How
Do People Rate?
Dynamic Models Amin Rahimian, Postdoctoral Associate, MIT Institute for Data,
Contributed Session Systems and Society, 100 Memorial Drive,, Apt 8-12A, Cambridge,
MA, 02142, United States, [email protected],
Chair: Amin Rahimian, MIT Institute for Data, Systems and Society,
Shrabastee Banerjee
100 Memorial Drive,, Apt 8-12A, Cambridge, MA, 02142,
United States, [email protected] Online reputation systems constitute an important component of many electronic
commerce platforms. Reviews and ratings posted on these platforms strongly
1 - Predicting Dynamic Consumer Search with influence the purchase and browsing behavior of customers, and play an
Reinforcement Learning important role in revenue generation. Existing literatures in marketing, computer
David M. Muir, Assistant Professor of Marketing, University of science, and economics highlight the statistical foundations of such impacts and
Delaware, Department of Business Administration, 20 Orchard Rd, address some of the behavioral and structural patterns that govern them (such as
Newark, DE, 19716, United States, [email protected] self-selection and reporting bias). These studies, however, fall short of a
comprehensive framework to capture all aspects of online reputation dynamics
In the process of consumer decision making, which frequently involves a that include the public perception of reviews, as well as the purchase and rating
thorough search of price and product information, consumers learn. Previous decisions. In this work, we propose a dynamic structural model for the evolution
research has considered the question of consumer search with learning from the of reputation on an online e-commerce platform. We use the proposed generative
perspective of causal inference; for example, previous work has assumed model to make inferences about individuals’ rating behavior. We train a deep
consumers use some form of a Bayesian updating process to learn as they search learning inference architecture with the simulated data from our generative
dynamically. Uniquely, I take a different approach to the study of dynamic search model to test for non-zero thresholds in the model. Finally, we apply the trained
with learning and apply the machine learning technique of reinforcement network to real time-series data of reviews submitted on a popular e-commerce
learning on clickstream data to show that both model-based MDP and the less platform. The results reveal circumstances under which these reviews are posted
restrictive, model-free Q-learning algorithm yield a complete set of predictions and their implications for the evolution of product ratings. Gaining a better
about how consumers search and learn in a dynamic world. Importantly, I understanding of the dynamics of reputation systems, namely, the conditions
employ the same dataset as previous consumer search research — which focus on under which ratings are submitted, and how they are eventually interpreted, is
causal inference instead of predictive accuracy - to illustrate how our respective crucial for marketers and designers of digital platforms, who can leverage this
works complement each other. This research has important practical implications information to stimulate further reviews and better manage user-generated
for marketers who want to encourage optimal real-time search pathways for content.
consumers.
2 - Building Hierarchy in the Dynamics of Marketing Mix Variables
Keyvan Dehmamy, Assistant Professor of Marketing, University of
Groningen, Nettelbosje 2, Groningen, 9747 AV, Netherlands,
[email protected], Jaap Wieringa
In the world of big-data, a familiar way to reduce the dimensionality could be
using factor analysis. This method however, often lacks structure and good
interpretability. One way to bring in structure to the extracted factors is to form
groups and apply a subjective hierarchy to the factors. The goal of this paper is to
identify the common dynamics of the marketing mix variables (like price or
advertisement activities) in the brands, category, and chain and market level. We
use an extension of a Dynamic Hierarchical Factor Models, in short DHFM
(Moench et al., 2013). This method was developed to add time varying intercepts
to a hierarchical factor model. To get an insight about possible common dynamics,
we use the information about two marketing mix variables (price and
advertisement activities) for one category of FMCG, including multiple brands
sold in several stores of different super-market chains in the Netherlands. We
extracted the common dynamics of the marketing mix variables into brand and
45
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 46
n FA12
Based on an updated agency-theory model, we propose a relationship between
cross-ownership and firm performance, and moderators of this relationship based
on the firm’s marketing capabilities and the firm’s strategic emphasis of marketing
Room 745, Alter Hall investment. To develop an industry cross-ownership measure, we use information
Marketing Strategy - Competition from 59,222 institutional owner individual holdings and then empirically test our
model using Arrellano-Bond estimation, accounting for endogeneity by using
Contributed Session instruments calculated using lags of the focal variables. We find a significant
Chair: Mitsukuni Nishida, Johns Hopkins Carey Business School, positive relationship between cross-ownership and firm performance over both
100 International Drive, Baltimore, MD, 21209, United States, the long- and short-term. Further, we find evidence that this effect is stronger
over both the short- and long-term for firms with lower marketing capabilities,
[email protected] suggesting that the effect of institutional cross-ownership on firm performance is
1 - The R2m Index Assessing the Relevance of Marketing stronger for firms with lower marketing capability. We also find that, over the
Scholarship to Business Practice short-term, institutional cross-ownership has a stronger effect on firms with a
Kamel Jedidi, John A. Howard Professor of Business, Columbia strategic emphasis towards R&D. Further, we find these results are robust to a
University, 518 Uris Hall Grad Sch of Bus, 3022 Broadway, New variety of alternative model and measure specifications. This framework and
York, NY, 10027, United States, [email protected], Bernd Schmitt, empirical results provide important managerial, public policy, and theoretical
implications on the consequences of cross-ownership.
Yanyan Li, Malek Ben Sliman
Has marketing academia strayed away too far from marketing practice and thus 4 - Firm Dynamics, Entry Order, and Performance
become less relevant? Many feel that way and have called for more emphasis on Mitsukuni Nishida, Johns Hopkins Carey Business School,
addressing relevant marketing problems. We present the development of a metric 100 International Drive, Baltimore, MD, 21209, United States,
that assesses the relevance of academic marketing articles to the practice of [email protected]
marketing. Ideally, academic articles should be evaluated directly by practitioners
Over the past forty years, a large literature on entry-order effects on performance
in terms of their relevance to practice to provide feedback to academics. However,
has documented how a pioneering firm entering a market or developing a
such a task would be tedious and time consuming. Instead, we propose the use of
product might (or might not) achieve an advantage in market share in reduced-
Natural Language Processing and Latent Dirichlet Allocation to classify articles
form analyses where they abstract away forward-looking decisions under strategic
into topics, and score each of the topics based on their “marketing-ness”.
interactions, which are often salient features of most industries. We empirically
Marketing terms are identified based on a marketing dictionary developed using
illustrate the importance of accounting for strategic interactions among firms and
various sources (e.g., AMA dictionary of marketing terms) and validated with
forward-looking incentive when measuring the entry-order effects on
practitioners. The proposed R2M (Relevance to Marketing) score is a weighted
performance. By using a panel on the store counts and sales from the
combination of the probability that the article belongs to a topic multiplied by the
convenience-store industry in Japan, we find evidence of a revenue and entry
“marketing-ness” of the topic. We apply the proposed R2M measurement on a
cost advantage for early entrants. In contrast, early entrants face a disadvantage in
balanced, random sample of more than 4000 articles from Journal of Consumer
expansion and variable costs.
Research (JCR), Journal of Marketing (JM), Journal of Marketing Research
(JMR), and Marketing Science from 1982 to 2015. We discuss the marketing
topics generated from the LDA analysis, show their evolution over time and assess
their “marketing-ness.” We also report the R2M score distribution overall and by
journal. We find that JM has the highest R2M score followed by Marketing
n FA13
Science, JMR, and JCR. For validity check, we find that the R2M measure Room 746, Alter Hall
correlates well with observable measures of relevance, such as practice prize
awards and self-stated evaluations by marketing practitioners of a random set of Multi-Sided Platforms 1
articles. Estimating the R2M on a holdout set of articles from MSI working paper Contributed Session
series, Psychological Review (PR), and Quarterly Journal of Economics (QJE), we
find that MSI working papers score slightly higher than JM; PR and QJE generally Chair: Jinzhao Du, Fuqua Business School, Apt 506, 910 Constitution
score lower than marketing journals as expected. Dr., Durham, NC, 27705, United States, [email protected]
2 - Technological Dynamism and the Janus Like Effects of 1 - Preference Elicitation, Price Discrimination and Matching in
Firm Capabilities Ride-sharing Systems
Saeed Janani, Arizona State University, BA 411, Tempe, AZ, Mustafa Dogan, Post-doctoral Associate, Carnegie Mellon
85281, United States, [email protected], Gaia Rubera, University, 5000 Forbes Avenue, Pittsburgh, PA, 15213,
Michael Wiles United States, [email protected], Alexandre Jacquillat,
Vibhanshu Abhishek
Technology is a key aspect of any markets, and markets go through various
technology phases, each of which rewards firms with particular set of capabilities. We study the dynamic design of a ride sharing platform in terms of allocation and
Having a contingency view on firm capabilities, we investigate the effects of pricing, when there is demand heterogeneity over the time preferences. There are
marketing, design, and R&D capabilities on return on assets (ROA) in different two type of agents: time-sensitive agents who value low wait times even at a price
technological market conditions. Using all the published patents in the US as of premium, and time-sensitive agents, with great flexibility regarding wait times.
1980, we defined three fundamental technology characteristics of markets: By allowing the agents to reveal their preferences and designing dynamic
technology intensity, technological growth, and technological uncertainty. While allocation and pricing rules accordingly, the platform can get an opportunity to
technology intensity refers to the level of technological advancement in a market discriminate over the customers, and, at the same time, smooth out the imbalance
in a certain year, technological growth and uncertainty respectively capture the between the demand and supply over time. We show that, when the valuation
general trend and the degree of unpredictability in technological changes in a heterogeneity is strong, time becomes a significant discrimination tool. In this
market. Building on the RBV and dynamic capability theories and using the most case, the platform can extract all the surplus that the system generates. This
comprehensive sample to date with 1961 firms within 378 industries (SIC codes) outcome however, is suboptimal as it does not maximize the total surplus. As
in 36 years, we manifest that all the capabilities have a positive effect on ROA, valuation heterogeneity becomes weaker, time becomes less important for
but technological conditions moderate such effects. That is, technology intensity discrimination, and the platform uses the allocation mechanism primarily to
has a positive moderating effect on the marketing capability-ROA relationship manage the dynamic imbalance between supply and demand. In this case, the
and a negative effect on the R&D capability-ROA relationship; technological platform has to leave some of the surplus to the time-sensitive agents as
uncertainty positively moderates the effects of these capabilities on ROA; and, information rent. We also show that the price does not necessarily increase with
technological growth has a positive impact on the R&D capability-ROA the realized demand, in contrast to conventional wisdom. We quantify the
relationship. This study provides insights to researchers and managers with benefits of our mechanism by comparing its performance to that of baseline
respect to what capabilities should be developed under different technological mechanisms based on surge pricing. We characterize the circumstances under
market conditions to drive profitability. which the general mechanism we suggest leads to more preferable outcomes in
terms of consumer surplus as well as the overall system efficiency.
3 - What if Your Owners Own Other Firms in Your Industry?
Institutional Cross-ownership’s Effect on Firm Marketing
and Performance
John Healey, Tulane University, 7 McAlister Drive, New Orleans,
LA, 70118, United States, [email protected], Ofer Mintz
The growth in institutional holdings of public firms has led to increased incidence
of cross-ownership, in which the same investor owns stakes in multiple firms
within an industry. Economic theory suggests such cross-ownership could impact
individual firm capabilities, but no empirical study has examined its association
with a capability like marketing that all firms need in order to sustain profitability.
46
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 47
3 - Media Platforms’ Content Provision Strategy and Source 3 - Assortment Optimization: Using Network Analysis to Reduce
of Profits Product Complexity at Retail Stores
Jinzhao Du, Fuqua Business School, Apt 506, 910 Constitution Ehsan Gholami, University of California, Davis, 690 Alvarado Ave.,
Dr., Durham, NC, 27705, United States, [email protected], Apt 11, Davis, CA, 95616, United States, [email protected],
Wilfred Amaldoss, Woochoel Shin Kevin Manuel, Ashwin Aravindakshan, Chen-Nee Chuah
We see media platforms earning their profits from both consumers and advertisers Assortment rationalization, the process to determine if a product should be kept
(e.g., the New York Times), advertisers only (e.g., the Huffington Post), or or discontinued, has gained renewed importance in the retail industry. Prior
consumers only (e.g., Tidal). This paper theoretically investigates two important research has produced several assortment rationalization algorithms. Nonetheless,
strategic issues confronting a media platform: what proportion of its limited to the best of our knowledge, none of the proposed solutions are comprehensive
bandwidth or space should a platform allocate for content (instead of when the number of Stock-Keeping Units (SKUs) increases into the thousands. In
advertising)? and what should be the source of a platform’s profits? To facilitate this research, we propose a graph-based approach in which the components of
this analysis, we propose a model where a media platform interacts with three the graph (e.g., nodes, edges) are constructed using the retail customer’s
sides: content suppliers, consumers, and advertisers. In a perfectly competitive purchasing history. This construction allows retailers to learn not only about SKUs
content market, our analysis shows that competing platforms will adopt a free- with high sales but also about key products that might bring shoppers into the
content strategy even in circumstances where a monopoly platform adopts a store. Given the vast number of SKUs (over tens of thousands of items) that each
paid-content-with-ads strategy. However, the result can get reversed if the retailer carries and the shelf space capacity constraint, we take advantage of graph
content supplier is a monopoly. Counter to conventional wisdom, inter-platform theory to reduce the number of products in an assortment. We do so by analyzing
competition helps a platform to earn more profits when they adopt a free-content the purchasing behavior of all customers to determine the relative importance of
strategy. Next, despite paying a lower price to content suppliers, a media platform different goods in the assortment. We find that our solution outperforms other
may still get hurt. Furthermore, though advertisers’ higher valuation for assortment rationalization methodologies. Finally, we also perform an analogous
consumers benefits a media platform, it can hurt a content supplier’s profits when analysis at the personal level, targeting the key merchandise in a shopping basket.
a monopoly supplier sells content to a platform using paid-content-with-ads Leveraging the results of the aforementioned analyses, we build a network of
strategy or when duopoly suppliers can shape consumers’ preference at a low relationships among products and develop a network-based algorithm to cull the
marginal cost and sell to a platform using free-content strategy. Finally, if assortment subject to the capacity constraint.
advertising is quite annoying, even while a monopoly platform shuns the 4 - Generalizing Benford’s Law for Retail Prices
advertising market, duopoly platforms may cater to advertisers. Vinay Kanetkar, Associate Professor, University of Guelph, Dept of
Marketing and Consumer Studies, University of Guelph, Guelph,
ON, N1G 2W1, Canada, [email protected]
n FA14 Newcomb (1881) observed that beginning pages of logarithmic tables were
overused than back pages. Benford (1938) independently noted for 20 different
Room 607, Alter Hall datasets that the first significant digit (FSD) did not have uniform distribution as
Machine Learning - Retail might be expected. Benford (1938) suggested that FSD distribution followed
logarithm of digit with base 10. Benford’s observations were empirical
Contributed Session generalization. Seven years later, Stigler (1945) proposed alternative explanation
Chair: Vinay Kanetkar, University of Guelph, Dept of Marketing and for Benford law. Over the last 135 year, many mathematician, physicist and
Consumer Studies, Guelph, ON, N1G 2W1, Canada, statistician have tried to explain this phenomenon using variety of techniques,
methods and datasets. In area of business, Benford’s law is used for fraud
[email protected]
detection. However, number publications in marketing and economics are
1 - Model-based Marketing Insights from High-dimensional limited. In this paper, we investigate whether retail prices follow pattern
Purchase Data suggested by Benford law. We examined retail prices for automobiles, computer
Bruno Jacobs, Assistant Professor, University of Maryland, 3453 monitors, end of day stock market prices, house prices (list and selling prices),
Van Munching Hall, Robert H. Smith School of Business, College dish detergent prices based on scanner data as well as prices for spirits (wine,
Park, MD, 20742, United States, [email protected], whiskey, rum, gin, vodka). Not surprising that Benford’s law generally does work
for most of retail prices. Using several flexible distribution that are special case of
Bas Donkers, Dennis Fok
Benford’s or Stigler’s law, we are able describe various retail price datasets.
We introduce a novel method to obtain insight in purchase behavior from high- Although there is substantial literature in the marketing about price-ending,
dimensional purchase data, e.g. by identifying seasonality and dynamics in strategic aspect of pricing is hidden in the first digit. That is should retailer price its
purchase behavior. In addition to these high-level insights, the method can product and service for $1 or $2 and so on and not about $1.39 or $1.40. This
predict on the level of a shopping basket which products will be relevant. We paper provides insights about distribution of first digits using retail prices.
extend on previous work where LDA was used to model purchase behavior using
latent motivations in several ways: We allow for correlations between
motivations, estimate time effects such as seasonality, and allow for dynamics
47
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 48
48
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 49
49
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 50
50
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 51
51
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 52
n FB05
adoption. Customers prefer mobile channel mainly because there are more
discounts for the transactions conducted via the mobile channel. We also find a
significant lock-in effect of lifetime, i.e., customers with shorter PC lifetime on e-
Room 232, Alter Hall commerce platform are more likely to adopt the mobile channel. Compared to
Mobile, Algorithm, and Artificial Intelligence (AI) mobile channel, online pay learning process is more affected by customer-specific
Session III: Customer Welfare in Mobile Age factors, including geographical and membership factors. We find that customers in
first-tier cities or in the higher membership level are less likely to adopt online
General Session pay. Three sets of policy simulations are conducted to optimize bonus policies for
firms to promote customers’ mobile pay adoption. The simulation results show
Chair: Xueming Luo, Temple University, Philadelphia, PA, 19122, that reducing bonus amount while increasing bonus offering opportunities and
United States, [email protected] extending bonus policy durations are more effective strategies.
Co-Chair: Siliang Tong, Temple University, Philadelphia, PA, 19144,
United States, [email protected]
1 - Personalized Recommendation Algorithms Impact on Seller n FB06
Revenues in the Sharing Economy
Siliang Tong, Temple University, Temple University,
Room 234, Alter Hall
1801 Liacouras Walk, Philadelphia, PA, 19144, United States, Analysis of Markets III – Competitive Strategies
[email protected] Contributed Session
A central feature of the sharing economy is to use recommendation algorithms to
Chair: Yusan Lin, Penn State University, 301D Grubb, White Course
match buyers and sellers. However, it remains unclear whether and how
recommendation algorithms affect demand and supply of products in the sharing Apartments, University Park, PA, 16802, United States, [email protected]
economy. The authors explore this question by using a unique field data from a 1 - When Should Biopharmaceutical Rivals form Alliances
leading Asian home-cooked food mobile sharing platform with1.5 million Cexun (Jeff) Cai, Assistant Professor, Texas A&M University,
observations. To match peer sellers who cook dishes at their home kitchens with College Station, TX, United States, [email protected]
peer buyers in the same city who order take-out food, the app platform Arun Gopalakrishnan
implemented two algorithms: (1) review popularity recommendation (RPR) and
(2) botler personalization recommendation (BPR). Data analyses find that both Biopharmaceutical firms, when developing products for new markets, sometimes
RPR and BPR have significantly positive incremental effects on the seller’s have to decide between forming an R&D alliance with a potential rival versus
revenues, but through different mechanisms. RPR can improve sales revenues by competing on R&D. When such alliances make sense for firms, as well as
benefiting the seller’s future online reputation on the demand side, whereas BPR regulators looking to promote innovation that improves consumer welfare,
can improve revenues by motivating sellers to offer more new products on the remains far from clear. In this paper we use a game-theoretic framework to study
supply side. The results also find unintended consequences of recommendation the conditions under which firms should pursue R&D alliances, and the situations
algorithm designs: RPR has negative impact on new product offering, and BPR under which the alliance decision may conflict with consumers’ interests in
reduces subsequent consumer review rating. These findings are consistent with having access to life-saving medicines. First, we find that R&D costs in an alliance
the exploration-exploitation tradeoff account, in which RPR acts as an relative to in-house costs are an important driver of the alliance decision. Second,
exploitation algorithm since it encourages sellers to leverage existing firms with large asymmetries in their in-house R&D costs are less likely to form
competences, whereas BPR acts as an exploration algorithm as it rewards sellers an alliance. Third, as product-market competition intensifies, R&D alliances are
who are willing to take risks and introduce new products to satisfy buyers’ more likely to be pursued to alleviate competition rather than for efficiency gains
personal preference. The authors discuss implications for how to leverage in R&D. Fourth, and most strikingly, we show a sizable region of the parameter
recommendation algorithms and artificial intelligence for the prosper of sharing space in which firms would pursue an R&D alliance even though it would lower
economy platforms. the chances of a successful drug becoming available to consumers -a clear loss
of ex ante consumer welfare that regulators should be concerned with. These
2 - Empowering Patients using Smart Mobile Health Platforms: results provide a novel theoretical analysis of the biopharmaceutical industry for
Evidence from a Randomized Field Experiment regulators, firms, and researchers.
Beibei Li, Carnegie Mellon University, 5000 Forbes Ave,
2 - Consumer Social Sharing and Brand Competition
Hamburg Hall 3026, Pittsburgh, PA, 15213, United States,
Jane Gu, Associate Professor, University of Connecticut,
[email protected]
2100 Hillside Road, Unit 1041, Storrs, CT, 06269, United States,
With today’s technological advancements, mobile phones and wearable devices [email protected], Xinxin Li
have become extensions of an increasingly diffused and smart digital
infrastructure. In this paper, we examine the emerging mobile health (mHealth) This study investigates how the identity revelation consequence of consumers’
platform and its health and economic impacts on the outcomes of diabetes social sharing behaviors affects their product choice decisions and firm
patients. To do so, we partnered with a major mHealth firm that provides one of competition. Our analysis reveals two identity revelation mechanisms induced by
the largest mobile health app platforms in Asia, specializing in diabetes care, two types of consumer sharing activities that have different strategic influences on
together with the Office of Chronic Disease Management from the national market competition. First, consumers commonly publicize on social media their
Ministry of Health. We designed and implemented a randomized field experiment associations with products of distinct brand images, which has the effect of
based on 9,251 unique responses from 1,070 diabetes patients over a 15-month suggesting their identities to the public. Such identity suggestion through
period from May 1, 2015, to July 31, 2016. Our main findings show that adoption publicizing product association allows consumers to derive a social value in
of an mHealth platform by users has a statistically significant impact on reducing addition to the consumption value from buying a branded product. Identity
blood glucose and glycated hemoglobin levels, hospital visits, and medical suggestion affects firm profit through three effects: the positive identification
expenses of diabetes patients over time. In conjunction with patient self- effect, the negative misidentification effect, and the positive competition
management through the mHealth platform, we also find heterogeneous effects alleviation effect. We find that compared to the case of no social sharing, identity
between personalized and non-personalized messages. Interestingly, non- suggestion through publicizing product association always enhances the profit of
personalized mobile messages with general diabetes-care guidance demonstrate a the leading firm in a market with a large loyal segment, but can hurt the profit of
stronger impact on patient health improvement. Our findings indicate the the follower firm with a small loyal segment. Moreover, the follower firm may be
potential value of mHealth technologies, as well as the importance of mHealth further worse off when it attempts to enhance competitive status through
platform design in achieving better healthcare outcomes. expanding its loyal segment. Second, in addition to publicizing product
associations, consumers can also publicize their opinions on social media through
3 - A Dynamic Structural Model of Customer Learning on Mobile Pay publishing blog posts and comments, which has the effect of directly announcing
Shaohui Wu, Tsinghua Uiversity, School of Economics their identities to the public. Such identity announcement through publicizing
and Management, Beijing, 100084, China, opinions allows consumers to derive a social utility even if they do not buy any
[email protected] products. We find that when identity announcement has low effectiveness, it
inhibits the positive identification effect of identity suggestion, making the
As one of the prominent components of Financial Technology (FinTech), the follower firm with a small loyal segment more likely to suffer a profit loss from
method of mobile pay is gradually accepted by more and more customers. This consumer social sharing. When identity announcement has high effectiveness, it
study develops a stochastic model to investigate the dynamics of individuals’ not only inhibits the positive identification effect of identity suggestion, but
mobile pay behavior. A Bayesian learning model is proposed and calibrated using eliminates the negative misidentification effect.
a rich data set. We find that there are two significant learning processes on both
mobile pay channel and online pay type. These two learning processes are
affected by consumer-specific factors, order-specific factors, and promotions from
firms. We find a significant discounts effect on customers’ mobile channel
52
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 53
n FB07
Room 237, Alter Hall
n FB08
Sharing Economy II Room 238, Alter Hall
Contributed Session Consumer Choice Models 1
Chair: Chenchen Di, University of Illinois, Urban-Champaign, Contributed Session
IL., [email protected] Chair: Joseph Jason Bell, University of Iowa, 382 Westgate Street,
1 - The Impact of Ride-sharing Services on New Car Demand: Apt 5, Iowa City, IA, 52246, United States, [email protected]
An Empirical Analysis of the US Market 1 - Form + Function: Aesthetic Product Design via Adaptive,
Isamar Troncoso Cortez, PhD Student, University of Southern Geometrized Conjoint
California, Los Angeles, CA, United States, Fred M Feinberg, Handleman Professor of Marketing and Statistics,
[email protected], Sivaramakrishnan Siddarth, University of Michigan, 701 Tappan St., Room R5324, Ann Arbor,
Jorge Mario Silva-Risso MI, 48109, United States, [email protected], Namwoo Kang,
Ridesharing services, led by Uber and Lyft in the US market, are one of the most Max Yi Ren, Panos Papalambros
visible and successful sectors of the sharing economy. Many financial analysts find Some “attributes” critical to product design are notoriously elusive. Conjoint
their rapid growth and future prospects to pose a serious threat to mainline shines in calibrating the importance of concrete elements like price, MPG, and
automakers, though, to our knowledge, no documented empirical evidence warranty length, but struggles with perceptual ones like “style” and “design”.
directly supports this claim. To fill this gap in the literature, we analyze new car These holistic, visually-conveyed attributes trade-off against traditional ones in
transaction data at the product segment level, from more than 1,200 car dealers consumers’ minds, so are critical to eventual market choices. Getting them right
in 68 different geographical markets in the US, spanning a five year period during consumes vast R&D and consumer research resources; yet current preference
which ridesharing services - led mostly by Uber- were gradually introduced in elicitation methods cannot reliably assess them. Here, we leverage real-time
different markets. We supplement this with Google Trends data on market- adaptive rendering — via a parameterized geometric model — to measure this
specific ride-sharing usage intensity. Together, these data provide a rich variation interplay between key design and traditional attributes in the passenger vehicle
in the treatment effect across markets, which enables us to identify the impact of market. The proposed “bi-level adaptive conjoint” method uses crowd-sourced
ride-sharing entry on new car demand at the product segment level. We estimate data to both measure vehicle style preferences based on 3D geometrization and
a fixed effects model on the number of transactions for each product segment at reveal tradeoffs between design/style attributes and functional ones (e.g., price
each dealer to identify the difference-in-difference coefficient associated with the and fuel efficiency). Bayesian part-worth estimation, online training via ranking
impact of UberX entry on its sales. We find that ridesharing services resulted in a SVM, and a crowd-based experiment allow the proposed method to pinpoint
two percent decline in the overall demand for new cars, and that the impact which styling elements differentially drive individual-level consumer reaction.
varied by product segment. The entry-compact segment was affected the most,
with sales declining by almost eight percent, which is consistent with the notion 2 - An Empirical Model of Screening Rule Choice
that the primary users are millennials in urban areas for whom ridesharing is a Joseph Jason Bell, University of Iowa, 382 Westgate Street, Apt 5,
realistic alternative to owning a car. This idea is further reinforced by the finding Iowa City, IA, 52246, United States, [email protected],
that other car segments, such as luxury, luxury SUV, and trucks, show little or no Sanghak Lee, Gary J Russell
impact.
Consideration sets contain a wealth of information about competition between
2 - The Bright Side of Waiting in a Peer-to-peer Sharing Market, alternatives and brands. Consideration set formation is often thought to proceed
Despite Being Mostly Dark using non-compensatory rules, which very often include discontinuities. Because
Chuang Tang, National University of Singapore, Biz 2 #B1-03, of this, modeling consideration sets can be chal- lenging. In this essay, we
1 Business Link, Singapore, 117592, Singapore, overcome this challenge with a model where consumers choose from among a set
of rules, and evaluate the rules by anticipating their impact. Our econometric
[email protected], Dai Yao, Junhong Chu
specification allows us to simulate the screening behavior of consumers,
Peer-to-peer (P2P) online sharing markets, which enable consumers to exchange understand patterns of brand competi- tion, and perform counterfactuals. We
their unused resources, such as cars and apartments, with their peers for a rental, estimate the model with Bayesian techniques using a unique dataset from the
are gaining momentum. Previous studies have examined the impact of the design automobile industry where considerations sets are observed.
of the markets and the observable attributes of the traders, such as gender and
age. In this paper, we focus on the impact of a new factor—the multi-stage
53
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 54
n FB09
priced differently. To the extent that this pricing strategy reduces credit rationing
and improves financial institution profitability, price discrimination based on risk
may be advantageous. Our research aims to determine the value of price
Room 239, Alter Hall discrimination based on customer risk. The current research surrounding risk
Pricing Strategies I based pricing is scant and empirical investigations exploring the topic are even
more limited (Edelberg 2006). To our knowledge, there is no empirical
Contributed Session investigation of the effects of risk based pricing that models (1) the financial
Chair: Praveen K. Kopalle, Dartmouth College, 100 Tuck Hall, institution’s decision to approve a loan (2) the customer’s decision to accept a
Tuck School of Business, Hanover, NH, 03755, United States, loan (3) the customer’s decision to default and (4) the agent’s decision to select an
offer for presentation to customers. Our results suggest that price sensitivity
[email protected] decreases with higher levels of customer risk. This difference in willingness to pay
1 - The Risk Signal of Price Among Complementary Products: provides the financial institution with suitable conditions for risk based pricing.
Evidence from Innovative Short-term Insurances Furthermore, we find that price discrimination based on risk will lead to double-
Jochen Reiner, Goethe University Frankfurt, digit increases in profitability and increase access to credit for lower-credit quality
Theodor-W.-Adorno-Platz 4, Frankfurt, 60323, Germany, borrowers.
[email protected], Julia Wamsler, Martin Natter 4 - Can’t Take the Heat? Field Experiment in Residential
The authors propose and investigate the risk-revealing role of price in the context Energy Conservation
of complementary products, such as optional insurances (e.g., short-term Praveen K. Kopalle, Professor of Marketing, Dartmouth College,
insurances, product insurances or extended warranties). We argue that 100 Tuck Hall, Tuck School of Business, Hanover, NH, 03755,
consumers use the price of an optional insurance as a cue to infer the risk United States, [email protected], Jesse Burkhardt,
associated with the product, service, or activity to be insured (i.e., core product). Kenneth Gillingham
Thus, high prices for optional insurances should be interpreted as a risk signal and
should negatively influence the choice probability of the core product. Yet, the In this study, we examine the results of a field experiment on minute level and
observed high prices for optional insurances suggest that it is common practice to appliance specific electricity consumption. The experiment took place during
neglect dependencies between the choice of the core product and the price of the critical peak load pricing days in 2013-2014 in a neighborhood in Austin, Texas.
optional insurance. In exploring the risk-revealing role of price, we address this The study includes five treatment groups: (i) installation of Nest Thermostat, (ii)
pricing problem and contribute to research on the allocative and informational setting up an online account to track energy consumption at the appliance level,
role of price. We test our conceptual framework with a series of experimental (iii) sending a text message, (iv) sending an actionable text message, and finally,
studies, a choice-based conjoint study, and a simulation study on innovative (v) sending a text message that has peak load pricing information. The
short-term insurances. The key results demonstrate the existence of the risk- households were randomly assigned to each of the five treatment groups and one
revealing role of price and its effect on consumer behavior. We find that an control group. We ran a triple differences model and included household and day
increasing insurance price not only negatively affects purchases of the optional fixed effects. Using data at the appliance-minute-level, we estimate a price
insurance but also of the core product. Our results further show that the risk elasticity of electricity demand of -0.17, and find that over 60 percent of this
signal emerges from the insurance price and not from the mere presence of an response can be attributed to air conditioning. The results of the field experiment
optional insurance in a purchase situation. Finally, we explore the properties of are also consistent with more recent research at the intersection of marketing and
the risk signal (i.e., its boundaries and strengths) and discuss consequences for neuro-science.
optimal pricing policies. We show that the current industry practice involves a
conflict of interest if the product to be insured and the optional insurance are sold
by different suppliers.
n FB10
2 - The Impact of Prior Price Dispersion and Basket Size on the
Effectiveness of Retailer’s Obfuscation Strategies Room 605, Alter Hall
Saurabh Bhattacharya, Lecturer, Newcastle University Business Customer Relationship Management I
School, 5 Barrack Road, Newcastle, NE1 4se, United Kingdom,
[email protected], Matthew Gorton
Contributed Session
Obfuscation is a strategy used by retailers to decrease the price sensitivity of Chair: Guilherme Bucco, Dona Gabriela, 263, Ap.303, Porto Alegre,
consumers, by increasing their search costs, but in a way in which does not Brazil, [email protected]
damage the reputation of the firm. Typical examples include providing different 1 - Dependence of Mere-measurement Effects on Past Experiences:
sizes of packaging for the same product categories across different stores, charging An Empirical Investigation in Financial Service Industry
different delivery options for internet purchases etc. Prior research establishes that Minjee Sun, University of Toronto, Rotman School of
such pricing strategies limit the use of price learning mechanisms by consumers.
Management, 105 St. George st, Toronto, ON, M5S 3E6, Canada,
With regard to price learning, extant research in obfuscation has, however, not
considered the importance of the degree of past price dispersion in the product [email protected], Mengze Shi, Yupin Yang,
category in which the obfuscation strategy is applied. We propose that the lower Xubing Zhang
the previous price dispersion in a product category, the higher will be the price Conducting surveys to collect customer feedbacks may provide firms
learning for a consumer, which will decrease the probability of consumers an unintended benefit: mere participation of customers in the survey can change
purchasing a brand when a retailer applies an obfuscation strategy to mask higher their subsequent behaviors. This effect is commonly referred to as mere-
per unit prices. Additionally, we propose that the relationship between price measurement effect (or, survey participation effect). A widely accepted
dispersion and learning is moderated by the consumer’s basket size. Such that explanation for the effect is self-generated validity, which states that completing a
even when the dispersion in prices is low, consumers’ price learning may still be customer-satisfaction survey will increase the accessibility of previous experience
weak, increasing the likelihood of purchasing a product when an obfuscation and lead to subsequent behaviors consistent with the answer. However, the
strategy is employed. Based on AC Nielsen consumer panel and retail scanner literature has neglected how the valence of previous experience may affect the
data for a large US market, we demonstrate empirical support for our proposed direction of mere-measurement effect, in particular, the mere-measurement effect
model in the carbonated beverage product category where obfuscation strategy is under negative product performance remains unexplored. In this research, we
common. Managerial implications are discussed. investigate the impact of product performance on the mere-measurement effect
3 - The Impact of Discriminatory Pricing Based on Consumer Risk: using a unique dataset from a financial securities firm. The data contains a unique
natural experiment in which a random sample of customers participated in
An Empirical Investigation using Indirect Lending Through satisfaction surveys, and the fluctuation in product experience (i.e. stock portfolio
Retail Networks performance) is observable. We estimate the mere-measurement effect and the
Christopher Amaral, PhD Candidate in Marketing, Queen’s dependence of mere-measurement effect on customers’ past portfolio
University, 143 Union Street, Kingston, ON, K7L 3N6, Canada, performance. Taking several issues of the data into consideration, we apply Zero-
[email protected], Ceren Kolsarici, Mikhail Nediak inflated Negative Binomial models and sample matching techniques. Our results
show that the mere-measurement effect indeed depends on the previous portfolio
There is strong evidence in the marketing literature that suggests that consumers
performance, and the choice of measures (e.g., the activeness of investments, the
often differ in their willingness to pay (Besanko, Dubé, and Gupta 2003),
size of investments, and retention).
highlighting the potential for firms to profit by varying prices by consumer
segment. While segments can be formed using various factors, one potential
factor that has received less academic attention in the price discrimination
literature is risk. The concept of risk is especially applicable in the context of
consumer credit, where lenders can classify borrowers into risk segments that are
54
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 55
55
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 56
56
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 57
57
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 58
n FB15
payment decision. Our research offers a number of novel findings that are useful
to practitioners in for- and non-profit businesses who are interested in PWYW
pricing. Furthermore, our research suggests that signals of trust placed on
Room 603, Alter Hall customers by a firm could result in enhanced reciprocal behavior; this implies that
CB - Perceptions & Choice perceived trust could be used as a novel marketing tool, with further managerial
implications.
Contributed Session
4 - Factors Moderating the Effectiveness of In-store
Chair: Hari Ravella, Virginia Tech, Blacksburg, VA, 24061, Mobile Promotions
United States, [email protected] Hari Ravella, Virginia Tech, 7054 Haycock Road, Falls Church, VA,
1 - How Non-ownership Physical Possession Impacts 22043, United States, [email protected], Dipankar Chakravarti
Object Valuation
Mobile marketing has seen explosive growth in recent years (Shankar 2016).
Charan K. Bagga, Assistant Professor, University of Calgary, While the growing popularity of mobile communications testifies to its
2500 University Dr NW, Calgary, AB, T2N 1N4, Canada, marketplace impact, it raises important questions about the mechanisms that
[email protected], Neil Bendle, June Cotte drive consumer (shopper) response to in-store mobile promotions versus other
Physical possession is known to exert a positive effect on object valuation more traditional in-store promotion formats such as shelf coupons and end-of-
regardless of whether the object is legally owned. If physical possession affects aisle displays. The present paper argues that, in contrast to traditional in-store
object valuation independent of legal ownership, it raises questions with crucial promotions that tend to be passive, in-store mobile promotions tend to be
implications for marketing: How do the non-ownership physical possession types intrusive and demand the recipient’s attention. Mobile strategy research reported
of renting and borrowing impact valuation? Evidence from four experiments in Forbes (Conner 2013) suggests that 90% of text messages get read within 3
demonstrates that the valuation (i.e., willingness-to-pay) for rented objects is minutes of delivery, perhaps aided by an urge to access new information upon the
greater than the valuation for non-possessed or borrowed objects. Borrowed alerting ping. Notwithstanding its irritation invoking potential, a promotional
objects are found to be valued more than non-possessed objects only when the message that has penetrated the attention barrier then has significant potential to
knowledge available for the non-possessed objects is low. The authors find that precipitate purchase behaviors. We argue that relative to an equivalent in-store
psychological ownership mediates the relationship between valuation and non- promotional offer delivered in a traditional format, an in-store mobile promotion
ownership physical possession. Additionally, psychological ownership varies for has a significantly higher likelihood of deflecting consumers off planned purchase
different possession types (i.e., ownership, renting, and borrowing) as its routines and provoking purchases of alternative brands not previously part of
contributing routes (control, self-investment, and knowledge) operate differently their consideration sets. We also argue that such mobile promotion effects are
for each possession type. As further evidence of the psychological ownership likely to be larger for consumers who are lower (versus higher) on innate deal
based theoretical account, the authors show that rented objects are not valued proneness, as well as for hedonic versus utilitarian products. We test our
higher than non-possessed objects if the control or self-investment routes of predictions using consumers recruited from a web panel who participate in a
psychological ownership are suppressed. Finally, the authors examine the computer-simulated supermarket shopping task.
moderating influence of product hedonism-utilitarianism, and consumers’
tightwad-spendthrift tendency on the valuation of rented and borrowed objects.
The authors find that tightwad consumers have a greater willingness-to-pay than
spendthrifts for rented objects, but not for borrowed objects. Further, the nature n FB16
of the product (whether it is perceived as hedonic or utilitarian) does not affect Room 231, Alter Hall
the willingness-to-pay for rented and borrowed objects.
2018 ISMS/MSI Gary Lilien Practice Prize II
2 - The Effects of Bariatric Surgery on Delay Discounting
Modeling in Obesity General Session
Ratnalekha Viswanadham, PhD Candidate, INSEAD, Boulevard de Chair: John H Roberts, University of New South Wales, 6/61 Kirribilli
Constance, INSEAD, Fontainebleau, 77305, France, Avenue, Kirribilli, Sydney NSW, 2061, Australia, [email protected]
[email protected], Yann Cornil, Liane
Schmidt, Liane Schmidt, Christine Poitou, Pierre Chandon,
Michèle Chabert, Judith Aron-Wisnewsky, Karine Clément,
Hilke Plassmann, Hilke Plassmann
n FB17
We investigate through a delay discounting model-based approach whether Room 31, Alter Hall
bariatric surgery impacts patients’ self-control abilities, which may contribute to UGC IV - Online Reviews
the success of this weight loss intervention beyond modifying the digestive tract.
Lean controls and bariatric surgery obese candidates perform a delay-discounting Contributed Session
task while their brains were scanned under functional MRI with incentive- Chair: Lakshmi Vana, London Business School,
compatible and monetarily-equivalent food and monetary rewards. We fit from
106 Sutherland Avenue, London, NW9 2QP, United Kingdom,
neuroeconomics literature the “as-soon-as-possible” model, a single-parameter
delay discounting model derived from the hyperbolic model that accounts for not [email protected]
only present bias of delay discounting but also time invariance of intertemporal 1 - Review Helpfulness Product Type and Authenticity
choices (i.e. when both a smaller-sooner reward and larger-later reward are offset Can Uslay, Rutgers Business School, Newark, NJ, 08854, United
by the same amount of time). Results show, through discounting parameter States, [email protected], Sevincgul (Sev) Ulu,
comparisons, that pre-bariatric obese patients exhibit more delay discounting Sengun Yeniyurt
behavior for food than lean controls. Both groups exhibit no difference in
monetary rewards, and this behavior remains consistent after surgery for obese Consumers increasingly rely on online WOM and helpful reviews for their
patients. Discounting behavior differences between the groups for food rewards purchase decisions. This study examines the influence of product type and
diminish six months after surgery, which is evidence of an effect of the surgery on authenticity on perceived helpfulness. Analyses of seventeen-thousand
behavioral aspects of obesity. Results by the conference will include mediating observations of 40 -roducts under two gender-specific product categories suggest
effects of biological markers and neural correlates to gain a more comprehensive that product category (male versus female), valence, and review-type (verified or
understanding of domain and physiological differences in delay discounting. not) significantly affect review helpfulness. Building on research on selectivity
model and authenticity, we found that women utilize a higher threshold to find a
3 - Pay Now or Pay Later – The Role of Payment Time in Pay What review helpful, and they care more about review credibility than men do.
You Want Pricing Therefore, gender-specific product type is important to understand the nature of
Raghabendra KC, Cambridge Judge Business School, University of helpful votes. We also examine the interaction of rating and gender-specific
Cambridge, Cambridge, United Kingdom, [email protected], products on helpfulness, with results suggesting that women are more likely to
Vincent Mak, Elie Ofek vote for negative reviews than men. Verified-review received more helpful votes
for women products and it shows that female consumers have become more
Pay What You Want (PWYW) pricing has garnered substantial attention from discerning in their trust of reviews, and they expect to see evidence and clues of
academics, businesses, policy makers, and the media over the past decade. authenticity than men do. We used Negative Binominal Regression which allows
Existing research has unearthed various mechanisms at play in a PWYW us to manage the features of count data as well as address the heteroscedasticity
exchange and provided substantial information on when and how businesses can in linear regression and the over-dispersion problem in the Poisson regression
take advantage of a PWYW pricing strategy. What is yet to be investigated is how model.Implications: Findings confirm the importance of considering demographic
the impact of these mechanisms varies with the payment decision time. We focus variables when modeling information search behavior. First, findings highlight
on the difference between before and after the seller’s delivery of value to the that gender-specific product types and authenticity of a review contribute to
customer. We propose that, compared with when the payment decision is to be helpfulness. Customizing information presented by gender can provide benefits to
made before the delivery of value, when a seller suggests the payment decision be companies. Online retailers can use new offers tailored to gender-specific
made after the delivery of value, the customer perceives a higher amount of trust requirements to increase perceived helpfulness of reviews and the overall
placed on them by the seller. This increased perceived trust is then reciprocated by platform.
the customer with an increased payment for the received value. In a series of
experimental studies, including two field experiments, we find confirming
evidence that the customer’s reciprocal response is enhanced with a delayed
58
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 59
n FC01 n FC02
Room 32, Alter Hall Room 33, Alter Hall
Panel: Where Industry Meets Academia: Digital Economy VII: Leveraging Data from
Biopharma Marketing Science Research Needs Digitalization
in a Changing Environment General Session
Panel Session Chair: Eva Ascarza, Columbia Business School, New York, NY, 10027-
6902, United States, [email protected]
Moderator: George Chressanthis, Axtria, 315 E. State Street, Berkeley
Heights, NJ, 0, United States, [email protected] 1 - The Impact of Subscription Programs on Customers’
Purchase Behavior
Moderator: Murali K Mantrala, University of Missouri, Columbia, MO, Raghuram Iyengar, University of Pennsylvania, 3730 Walnut
65211, United States, [email protected] Street, 700 JMHH, Philadelphia, PA, 19104, United States,
1 - Where Industry Meets Academia: Biopharma Marketing Science [email protected], Young-Hoon Park, Qi Yu
Research Needs in a Changing Environment Firms use various methods to boost sales and to keep their customers alive.
George Chressanthis, Axtria, Berkeley Heights, NJ, United States, Recently, subscription programs have become increasingly popular among
[email protected] retailers including Amazon (Prime), Sephora (Flash), and Barnes & Noble (B&N
The biopharmaceutical industry has long been used by academicians as an Membership). This type of programs bundle a series of benefits, e.g. free shipping,
incubator for new marketing science innovations as evidenced by a small member-exclusive discounts, etc., for a fixed periodic fee. Despite the popularity
representative sample of the total studies performed references cited here. A few of subscription programs, their effects on sales and customer retention remain
reasons for this keen interest is not only the importance of this industry to overall unclear. In this research, we leverage a unique dataset from a fast moving
societal welfare but also the availability of granular data that allows for the consumer goods company to study the causal effect of customers’ adoption of
applications of new marketing science ideas that would be more difficult to subscription program on their subsequent purchases. The company launched a
perform in other industries. However, the biopharmaceutical industry has been standard subscription program in its online channel in 2015, allowing customers
undergoing dramatic shifts caused by changes in the external environment that to have coupons, free samples, member-exclusive promotions, free shipping,
are altering the metrics that define brand success and new developments in along with other benefits with an annual fee of $50. In order to evaluate the
scientific R&D that have moved drug portfolios toward focusing on large molecule causal impact of this program and to account for the fact that high-value
specialty medicines. For example, biologics account for more than 30% of total customers self-select into the program, we adopt a quasi-experiment setting. We
US drug spending, yet comprise less than 1% of all dispensed prescriptions. More first match each member with a non-member with similar demographics and
new drug applications are classified now as orphan drugs that cater to purchasing records and then estimate a difference-in-differences model on the
personalized targeted therapies and small patient populations. Brand success and matched sample. We find that customers on average increase their monthly
treatment paradigms as assessed by payers and providers are increasingly being purchases by $27 after join the subscription program. Moreover, there is
determined by measures of value such as improvements in health outcomes, substantial heterogeneity in the treatment effects. The subscription program is
overall treatment costs, and cost-effectiveness. These issues are especially acute in more effective on customers with higher initial purchases than those with lower
the area of new oncology drugs where there is significant focus by initial purchases. Customers also increase their consumption variety after joining
biopharmaceutical companies, given their higher costs/risks of R&D and problems the program, suggesting that customers switch their purchase from the firm’s
of affordability/access for payers and patients. This change in assessment has competitors and increase their share of wallet. We confirm this mechanism by
meant the inclusion of claims data and electronic health records to augment demonstrating that customers more willing to try new products experience a
traditional commercial data in the conduct of biopharmaceutical sales and larger increase in their purchases and consumption variety.
marketing analytics in order to achieve process outcomes (e.g., sales force 2 - The Value of First Impressions: Leveraging Acquisition Data for
optimization, marketing-mix, promotion-response modeling, etc.). Customer Management
Biopharmaceutical company commercial operation organizations have been
Nicolas Padilla, Columbia Business School, 3022 Broadway,
reshaping their sales and marketing processes in response to these and other
environmental changes. Yet, one commercial area that has lagged behind in Uris Hall, 5th Floor, New York, NY, 10027, United States,
response to these developments is new marketing science innovations that [email protected], Eva Ascarza
industry practitioners can apply to create solutions within this changing Firms increasingly have access to richer customer data. What a decade ago was
environment. Insights from industry executives from Bayer, Genentech, GSK, merely a transaction added to a customer database has become a collection of
Janssen, Merck, Novartis, Pfizer will be shared in this special panel session. behaviors that a customer engages in while she is making a purchase (e.g.,
whether her purchase was online or offline, whether she used a tablet or
computer, whether she bought a new product or an old best-seller). These data
can be used to form the ``first impressions’’ of a customer based on her initial
transaction. We posit that a customer’s first impressions carry valuable
information for the firm by explaining a large proportion of the heterogeneity in
future behavior, both in terms of what the customer is expected to do (i.e., her
lifetime value) and how responsive she will be to marketing actions. The latter
point is especially relevant for contexts in which firms do not observe many
purchases per individual (e.g., retail) or where targeting occurs soon after
customer acquisition. In these contexts, models that only rely on unobserved
sources of heterogeneity are unable to help the marketer target newly acquired
customers with precision. We develop a model incorporates latent dimensions
extracted from observed heterogeneity—capturing the effect of first impressions—
both in individual propensity to buy as well as responsiveness to the firm’s
marketing actions. Applying the model to data from a retail context, we show
that the focal firm would significantly improve the return on their marketing
actions if it targeted just-acquired customers based on their first impressions.
59
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 60
60
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 61
n FC05 With a growing number of branded mobile apps launched on the market,
marketers are facing fierce competition in attracting consumers to download their
apps. As such, many marketers have devoted increased marketing efforts to
Room 232, Alter Hall promote their apps. However, little is known about the effectiveness of mobile
Mobile, Algorithm, and Artificial Intelligence (AI) marketing strategies on consumers’ adoption of branded apps. In this study, we
Session IV: Recommendation Algorithm and investigate if three mobile marketing strategies−mobile advertising, mobile
regular couponing and mobile group-buying couponing−create different impacts
Deep Learning on consumers’ adoptions and how social influences in the virtual environment
General Session (i.e., the installed base of adopters) and in the physical environment (i.e., the
crowdedness in the mall) strengthen or weaken the effectiveness of three
Chair: Xueming Luo, Temple University, Philadelphia, PA, 19122, strategies. Based on a proprietary data collected from a large Chinese shopping
United States, [email protected] mall, our study shows that the adoption rate of its branded app is increased with a
large number of regular mobile coupons and group-buying coupons issued, but
Co-Chair: Kunpeng Zhang, University of Maryland, College Park, MD, decreased with a large number of mobile advertising issued. As the number of
75248-4055, United States, [email protected] adopters grows, we find that while the positive effect of mobile regular couponing
1 - Design of Fashion: Can Brand Value be Separated from becomes stronger, the positive effect of mobile group-buying couponing and the
Style Value? negative effect of mobile advertising become weaker. Most interestingly, the social
influences of people in the mall (i.e., the crowdedness) create opposite
Kannan Srinivasan, Carnegie Mellon University, Tepper School of
moderating effects on the effectiveness of three mobile marketing strategies.
Business, 5000 Forbes Avenue, Pittsburgh, PA, 15213, United Specifically, as the mall becomes crowded, the positive effect of mobile group-
States, [email protected], Zijun (June) Shi, Dokyun Lee, buying couponing and the negative effect of mobile advertising becomes stronger,
Param Vir Singh but the positive effect of mobile regular couponing become weaker.
In fashion, consumers value both the brand and the style of an item. On the
firms’ side, they strive to build up brand value and design fashion items for each
new season. However, little empirical study has been done on quantifying the
brand value, style value, and especially the substitutability between them in the
fashion market. In this research, we seek to understand 1) how brands are valued
in the fashion market, 2) how substitutable are brand and style features, and 3)
how to design a fashion look that attracts attention in the online community.To
do so, we collect a dataset from a major online fashion social networking
community. We first employ deep learning based computer vision techniques to
create style features, which quantify a fashion look’s styles, including clothing
61
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 62
62
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 63
n FC08
which runs in polynomial time, can be used to implement the randomized
algorithm. The maximum likelihood value determines a lower bound on the
performance ratio of the randomized algorithm. We employ the proposed
Room 238, Alter Hall approach to infer lexicographic rules for individuals using data from a choice
Consumer Choice Models II experiment for electronic tablets. These rules obtain substantially better fit and
predictions than a previously described greedy algorithm, a local-search
Contributed Session algorithm, and multinomial logit and probit models.
Chair: Rajeev Kohli, Columbia University, 506 Uris Hall, Graduate
School of Business, New York, NY, 10027-6902, United States,
[email protected] n FC09
1 - A Model of Multi-category Choice and Quantity Consumed
in Snacks Room 239, Alter Hall
Sriharsha Kamatham, UT Dallas, Richardson, TX, 75080, United Pricing Strategies II
States, [email protected], B.P.S. Murthi, Marina Girju Contributed Session
When snacking, consumers may eat a large variety of snacks within a day. We
develop and estimate a model of category choice and quantity consumed using Chair: Yu-Hung Chen, Doctor, National Taiwan University, Taipei,
individual-level snack consumption data. We use a model derived from random 10617, Taiwan, [email protected]
utility maximization theory that helps us understand variety seeking in a multi- 1 - Information Good Pricing and Consumer Deliberation
category consumption context. The model is derived from the multiple Shan-Yu Chou, Professor, National Taiwan University,
discrete-continuous framework proposed by Bhat (2005, 2008), Kim et al. (2002), Dept of Bus Admin, 1 Sec. 4 Roosevelt Road, Taipei, 106, Taiwan,
enabling us to study variety seeking using the nested structure of snack brands
[email protected], Chyi-Mei Chen, Chung-Ting Tsou
and product categories. In the model, we are able to estimate the effects of state
dependence and satiation with respect to calories consumed, in addition to the This paper intends to analyze the pricing mechanisms of information goods when
effect of individual characteristics and contextual variables. We find that different firms facing consumers with uncertain use frequency and need to incur a
product categories exhibit different levels of inertia and satiation. The model deliberation cost to discover their use frequency. There are two mechanisms,
suggests different mechanisms for targeting heavy snack consumers. selling, where up-front payment allows unrestricted use, and pay-per-use, where
payments are tailored to use. Following Balasubramanian et al. (2015), we
2 - Uncovering Shopping Goal Structure from Consumer assume that paying-per-use impose some disutility on consumers, termed as
Purchase Histories psychological cost. We show that adopting both selling and pay-per-use schemes
Yang Pan, University of Iowa, S252 John Pappajohn Business in a monopoly yields the highest profit when the deliberation cost is sufficiently
Building, Department of Marketing, Iowa City, IA, 52242, low and the psychological cost is sufficiently high. If the deliberation cost is not
United States, [email protected], Gary J Russell low enough, then the monopolistic firm will optimally use a single pricing
scheme, either selling only or pay-per-use only to induce impulse buying to avoid
Consumer behavior in a retail setting is goal-directed behavior. Motivational the costly consumer deliberation. In this situation, if consumer deliberation cost is
theories of social psychology suggest that shopping goals influence consumers’ high enough, selling scheme is preferred to pay-per-view scheme and optimal for
behaviors and actions. Shopping goals can be abstract or concrete under the firm in that consumers, facing selling scheme, will voluntarily make
deliberate or implemental mindset. Concrete shopping goals also vary across purchasing decisions without deliberation even without price discounts. On the
consumers. Overall, within the same retail setting, different consumers may vary other hand, if deliberation cost is not sufficiently high, then pay-per-view may
in their shopping goals, which will be reflected in consumers’ intrinsic category yield a higher profit than selling provided the associated psychological cost is low
preferences that can be identified from consumer shopping histories. In this enough. In the duopoly model, we show that when consumer deliberation cost is
paper, we study shopping goals in a specific retail setting: a convenience store low relative to psychological cost, then in equilibrium the two competing firms
chain primarily located in the Midwestern United States. We assume the basket can both make profits by differentiating their pricing schemes, one offering
choice within each shopping trip follows a Multivariate Logistic Model. From this, selling-only scheme and the other offering pay-per-view only. When consumer
we derive that the short-run basket counts within each week follow the Poisson deliberation cost relative to psychological cost is not low enough, the two firms
distribution. We then identify the short-run basket model by a two-step cannot soften price competition by employing different selling schemes. In
procedure. In the first step, we analytically aggregate the short-run basket model particular, the firm employing pay-per-view, facing the associated consumer
over time to obtain the long-run basket model, which also takes the form of psychological cost, cannot make profits.
Poisson. We use the long-run basket data to obtain a consistent estimation of
consumers’ category preferences, which provide us some insights into the 2 - Optimal Design for Pay-what-you-want Pricing
structures of concrete shopping goals. In the second step, we combine short-run Ranjit M. Christopher, Asst. Professor, University of Missouri -
basket data and the estimates from the first stage to identify the rest of the Kansas City, 4768 Oak St, Apt 518, Kansas City, MO, 64112,
parameters in the short-run model. We report on an application of the model to United States, [email protected],
consumer choice histories from the convenience store retailer.
Fernando Sobral Machado
3 - The Effect of Perceived Cost on Package Choice and Purchase Pay-What-You-Want (PWYW) is an interesting pricing mechanism where firms
Quantity Decisions relinquish pricing power entirely to consumers. Despite recent attention in the
Haruka Kozuka, PhD Student, Keio University, Minato-ku, Shiba, topic, the existing empirical research has lacked solid grounding on a theoretical
5-11-4, Tokyo, 1080014, Japan, [email protected] model of consumer choice and has provided partial and mixed evidence on the
conditions under which PWYW may be profitable. In this research, we employ a
The purpose of this study is to clarify how consumers’ perceived cost affects
three-pronged approach to address how firms can use managerially controllable
package choice and purchase quantity decisions. This paper assumes that unit cost
variables such as payment anonymity and price recommendations to optimize the
consumers perceived from package size and type is a determining factor of
design of PWYW pricing to maximize firm goals such as market penetration and
consumers’ decisions — “which” and “how many” packages they should
revenues. First, we propose a simple theoretical model to derive conditions under
purchase. Then the research proposes a choice model which extends Bhat’s
which consumers will opt-out of the PWYW exchange, free ride, or make a freely
(2005) choice model by incorporating the influence of unit cost. For model
chosen positive payment. Our model explicitly accounts for variations to both
estimation, this paper uses experimental data collected from consumers’ package
consumers’ ease of participating and their valuation of the good under a PWYW
choice of a certain brand with three different packages design. The result of model
exchange contingent on the type of design employed. Second, by specifying
comparison shows that the fitness of our choice model is better than a choice
consumers’ ease of participation and their valuation as latent variables and
model not incorporating the effect of unit cost. In the result of model estimates,
functions of both the reservation price of the good under fixed-pricing and design
we find that unit cost, such as the cost for inventory, quality deterioration and
variables, we derive a statistical estimation procedure to identify the parameters
shortage, influences on satiation of product utility. In addition, it indicates that
of the theoretical model. Third, using a controlled experiment, we generate a
consumers’ heterogeneity affects the impact of unit cost on decision making.
sample of consumer responses to a generic PWYW design in order to calibrate
4 - Randomized Algorithms for Lexicographic Inference model parameters, validate our theory, and compute optimal values for
Rajeev Kohli, Ira Leon Rennert Professor of Business, Columbia managerially controllable pricing design variables. Finally, we discuss how
University, 506 Uris Hall, Graduate School of Business, New York, managers can employ our approach to optimize PWYW pricing specific to a
product/market context.
NY, 10027-6902, United States, [email protected],
Khaled Boughanmi, Vikram Kohli
The inference of a lexicographic rule from paired comparisons, ranking or choice
data is a discrete optimization problem that generalizes the linear ordering
problem. We develop an approach to its solution using randomized algorithms.
First, we show that maximizing the expected value of a randomized solution is
equivalent to solving the lexicographic inference problem. As a result, the discrete
problem is transformed into a continuous and unconstrained nonlinear program
that can be solved, possibly only to a local optimum, using standard nonlinear
optimization methods. Second, we show that a maximum likelihood procedure,
63
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 64
n FC11 n FC12
Room 606, Alter Hall Room 745, Alter Hall
E-Commerce 2 Methods – Mediation, Moderation & Segmentation
Contributed Session Contributed Session
Chair: Yong Chin Tan, Singapore Management University, Chair: Spyros Zoumpoulis, INSEAD, 15 Rue Daubenton, Paris, 75005,
113 Lorong 1 Toa Payoh, #05-460, Singapore, 310113, Singapore, France, [email protected]
[email protected] 1 - Asymmetric Relations and the Friendship Paradox
1 - Effect of Freight Subsidies on High Value and Expensive to Malek Ben Sliman, Ph.D. Candidate, Graduate School of Business,
Ship Products Columbia University, 5j Uris Hall, Graduate School of Business,
Fan Yang, University of British Columbia, 2111, Lower Mall, New York, NY, 10027, United States,
Vancouver, BC, V6T 1Z4, Canada, [email protected], [email protected], Rajeev Kohli
Chunhua Wu, Charles B. Weinberg The friendship paradox says that your friends have more friends than you do. The
In practice, many firms use partitioned pricing strategy, listing both product price result relies on the symmetry of friendship - if I am your friend, then you are also
and other price component such as shipping fee, separately. For this study we my friend. We generalize the paradox to allow asymmetric (leader-follower)
focus on an online seller of building products and examine the impact of a freight relations, such as those between Twitter followers and their leaders. We show that
subsidy on clickstream behavior and on sales. One key issue is consumers’ a person has, on average, as many leaders as followers, and that this common
sensitivity to the two price components—the price of the product itself and to average is no greater than the average number of followers per leader and leaders
freight. As freight costs on average account for 17% ($260) of the product sales per follower. The result implies that, for example, the people you follow on
price, knowing the differential response, if any, to product promotions and freight Twitter are potentially more influential than you (because they have more
promotions is crucial. Studies of consumer products, e.g. the sale of online books, followers than you) and the people who follow you are more connected and
has found such a differential response, but that is for relatively low value potentially better informed than you (because they follow more people than
products. In this paper, we develop a structural model to estimate consumers’ you). In the friendship paradox, the difference between the average number of
sensitivity to the two different partitioned components using clickstream data friends of friends and the average number of friends increases with the variance
(including purchase data with detailed price and cost information) from an online in the number of friends. In the present generalization, the difference between
shopping platform. The company implemented two freight subsidy promotions the average number of followers per leader and the average number of followers
during the time of our study. Focusing on the results for freight subsidies, we find increases with the variance in the number of followers; and the difference
that shoppers are responsive to price promotions on freight in terms of conversion between the average number of leaders per follower and the average number of
rate, however the effects on search behavior and total revenue are more nuanced. leaders increases with the variance in the number of leaders. We analyze a sample
Managerial implications of these results are presented. of over 28 million Twitter users and 193 million followers and leaders. On
average, a person in the sample has about 19 followers and leaders each, 1,250
2 - Optimal Utilization of Daily Deal Promotion in followers per leader and 281,000 leaders per follower. We examine the
Competitive Environments distributions of the number of leaders, the number of followers, the leaders of
Seong Kyoung Shin, Purdue University, West Lafayette, IN, 63117, followers, and the followers of leaders. All four distributions have long tails. A
United States, [email protected], jia li log-normal distribution characterizes the number of leaders in the data. The
distributions for the number of followers and the average number of leaders per
Over the past years, daily deal platforms such as Groupon have been popular and follower are both bimodal. This is consistent with a large fraction of users having
become the important format to connect customers and local merchants that offer very few followers, a smaller fraction belonging to interest-based communities,
activities, goods and services. Although daily deal promotion generates and a still smaller fraction of users being celebrities with large numbers of
incremental revenue, it is harshly criticized due to deep discounts, cannibalization followers.
and high commission. However, plenty of merchants still use daily deals because
most local merchants not only have difficulties to advertise their business, but 2 - Latent Variable Moderation Analysis in Marketing Research
also to face intense competition. Previous literature investigates whether daily Constant Pieters, Tilburg University, Tilburg, Netherlands,
deals are beneficial to merchants or not, but a competitive environment is not [email protected], Rik Pieters, Aurelie Lemmens
taken into consideration. In this research, we would like to understand the value
of daily deals for local merchants in competitive markets based on transaction Moderation analysis is common in marketing. It examines whether the effect of
data of merchants in China. Specifically, we develop a model to figure out variable X on Y is conditional on another variable Z. At least one of the variables
whether daily deals can be helpful for merchants to compete with others and how X and Z is often a latent construct, measured with multiple items. In such “latent
merchants make the optimal decisions when they offer daily deals. Our data moderation models”, the challenge is to identify the underlying moderation
reveals whether customers redeem daily deal vouchers upon their visits, which effect, while correcting for measurement error in the items. Improperly
helps us examine the role of daily deals in competitive circumstance. Thus, we accounting for measurement error increases the potential Type M (magnitude)
can explore not only how merchants to utilize daily deals strategically, but also and Type II error. Our study examines the various methodologies used in
how daily deal platforms manage their business more successfully. marketing research to deal with latent moderation, and how well these
methodologies are able to capture the true underlying conditional effect. First, a
3 - Information Targeting in Customers’ Online Purchase Journey review of 98 articles published in Journal of Marketing and Journal of Marketing
Yong Chin Tan, Singapore Management University, Singapore, Research between 2000 and 2017 reveals that about 90% of hypothesized
Singapore, [email protected], Srinivas Reddy, moderation effects are tested with an “observed moderation” approach, which
Sandeep R. Chandukala does not account for measurement error. The remainder of the tests use a “latent
measurement moderation” approach or a “latent structural moderation”
Online retailers often provide additional information related to product popularity approach, which do account for measurement error. Second, typical data from the
(e.g., number of customers who viewed / purchased) or scarcity (e.g., number of reviewed articles is simulated for Monte Carlo simulations. These show that the
units available) to persuade customers to purchase. However, the impact of observed moderation approach attenuates the estimated moderation effect by
providing these information in an online environment is not well understood. For about 25%, whereas there is minimal bias for the latent moderation approaches.
example, it is unclear if these information should be presented in early stages of Ironically, the latent moderation approaches increase the multicollinearity
the decision-making process when customers engage in exploratory browsing, or between the predictors in the model. Whereas multicollinearity increases the
later stages after they have narrowed their consideration sets. In addition, the standard error of the main effects and moderation effect, which is well-known, it
impact of presenting these different types of information on customers’ online also increases the reliability of the moderation effect, which decreases the bias.
purchase journey have not been investigated. Using a series of field experiments, Unsurprisingly, multicollinearity decreases the power to find true main effects,
the current study examines how the type of information, the way it is presented, but surprisingly it also increases the power to find true moderation effects. These
and where it is presented in customers’ decision-making process influences results demonstrate the importance of controlling for measurement error in
website engagement, product consideration, purchase conversion, and post- moderation models: reducing estimation bias while maximizing statistical power
purchase product returns. The moderating effects of various product and to find the true moderation effects.
customer characteristics are also explored, allowing retailers to target information
presentation based on these characteristics.
65
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 66
66
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 67
67
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 68
68
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 69
69
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 70
70
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 71
71
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 72
72
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 73
n FD07 n FD08
Room 237, Alter Hall Room 238, Alter Hall
Social Media - Influencers & Motivations Consumer Choice Models III
Contributed Session Contributed Session
Chair: Nils Wloemert, Vienna University of Economics and Business, Chair: John H. Roberts, University of Sussex, Brighton, United
Institute for Interactive, Welthandelsplatz 1, Vienna, 1020, Australia, Kingdom, [email protected]; Songting Dong, Huy Nguyen
[email protected] 1 - Modeling Reference Dependence Effect and
1 - A Model for Detecting Influencers in Social Media Regulatory Orientation
Mirai Igarashi, 2nd grade in MA, Tohoku University, I-Hsuan Chiu, Assistant Professor, University of Wisconsin -
27-1, Kawauchi, Aoba-Ku, Sendai-Shi, 980-8576, Japan, Milwaukee, Milwaukee, WI, United States, [email protected],
[email protected], Nobuhiko Terui Gary J. Russell
In modern social media development, viral marketing, which aims at efficient
One major implication of the value function in prospect theory is loss aversion:
information diffusion through word-of-mouth by socially influential people, or
losses loom larger than gains of the same size. The empirical examinations of loss
influencers, is an important area of study. This article propose statistical models
aversion from consumer brand choice decisions, however, provide mixed results.
for detection of the communities formed on social media and extraction of topics
For example, Bell and Lattin (2002) show that loss aversion is reduced or
that people in the community are interested in. Because the proposed model
disappears when taking into account consumer heterogeneity. In this research,
detects influencers from social network and enables marketing in consideration of
we show that regulatory focus theory can account for the heterogeneity in the
topics that people are interested in, marketers can spread information on products
reference dependent effect: consumers in a promotion-oriented mindset show a
and brands more effectively. In empirical analysis using Twitter data, we find that
weaker response to loss aversion comparing to those in a prevention-oriented
one of the communities detected from network structure is a community that
mindset. We consider a decision context where consumers rely on external
collects links from most users despite the small number of users, that is, a
reference points. To capture the reference dependence effect, we use the random
community to which many influencers belong. In addition, we reveal topics that
regret minimization (RRM) model developed by Chorus (2012). The RRM model
users in those communities are interested in, such as latest game machine and
describes a decision process that compares attributes of an alternative to that of all
electronic devices.
other alternatives in the consideration set. RRM allows for flexibility in evaluating
2 - Motives for Gifting in Live Streaming reference dependence effects across all attributes and alternatives, and provides a
Xuejing Ma, Peking University, Beijing, China, framework for incorporating regulatory orientations into the model specification.
[email protected], Hongju Liu, Qiaowei Shen Using data from a choice experiment, we demonstrate that prospect theory
describes the within-individual differences with respect to losses and gains, while
Recently, live streaming has enjoyed a surge in popularity around the world. regulatory focus theory describes the between-individual differences.
Besides free engagements in live streaming such as watching and chatting,
viewers can also purchase virtual gifts and send them to broadcasters. Using data 2 - Identifying Pivotal Attributes for Choice of Variety
from one of the largest live streaming platforms in China, we empirically explore Sanghak Lee, Assistant Professor of Marketing, Arizona State
the motives that drive viewers to gift. We find evidence of crowding out effect, University, P.O. Box 874106, Tempe, AZ, 85287, United States,
i.e., a larger cumulative amount of gifting tends to reduce the probability that a [email protected], Sunghoon Kim
new viewer will gift after joining the session. We propose two alternative
explanations for the crowding out effect. On the one hand, viewers may send gifts Consumers’ market basket often contains more than one variety of a product that
to broadcasters driven by an image-related motivation. The larger the cumulative pursue the same primary purpose while serving differentially with differences in
amount of gifting, the more a viewer has to spend to get attention from selected attributes. The pivotal attributes for variety are not observable and may
broadcasters and other viewers. On the other hand, viewers may care about the vary across consumers. For example, consumers who seek for variety in flavor
total amount of gifting a broadcaster would receive, in which case the gifting of would buy multiple flavors together (e.g., strawberry yogurt and blueberry
others and that of themselves are substitutes. Further analysis reveals two yogurt), while others who consider different consumption situations would
segments of viewers. Conditional on gifting, viewers in one of the segments are include various package types (e.g., a 6oz multi pack and a 32oz single pack) in
likely to gift more when the cumulative amount of gifting from others is larger; his/her shopping basket. We develop a Bayesian choice model that accommodates
while for viewers in the other segment, the amount of gifting is negatively related demand for variety and identifies the pivotal attributes for variety-seeking. The
to the cumulative amount of gifting from others. proposed model relies on the satiation at the pivotal attributes as a source of
variety-seeking and allows for consumer heterogeneity in selecting the pivotal
3 - Online Content Monetization: How Paywalls Affect Human attributes for variety. The empirical support for our model is provided using a
Brand Success consumer panel dataset of carbonated beverage products, demonstrating its
Nils Wloemert, Assistant Professor, Vienna University of implication for cross-price elasticity.
Economics and Business, Institute for Interactive Marketing & 3 - A Cross-validity Comparison of Likelihood Methods for
Social Media, Welthandelsplatz 1, Vienna, 1020, Austria, Distributions with Intractable Normalizing Constants
[email protected], Christian Hotz-Behofsits, Tetyana Kosyakova, Frankfurt School of Finance & Management,
Nadia Abou Nabout Adickesallee 32-34, Frankfurt, 60322, Germany, [email protected]
The increasing use of social media enables influential users to market themselves Statistical models with likelihood functions that are difficult to evaluate or are
as human brands (e.g., YouTubers) by creating content (e.g., videos) that is intractable have received much attention in the last decades. This problem is
consumed by their followers. Many social media platforms allow human brands known in a wide range of contexts, e.g., in spatial econometrics, statistical
to monetize their content through advertising, or subscription fees, or both. genetics, or network analysis. Researchers in marketing encounter this problem,
Choosing the right monetization strategy is not a trivial task for human brands. for example, when modeling choices from menus, where demand for items is
While providing content free of charge may increase brand value by attracting a interdependent because of substitution and complementarity relationships.
large audience, charging for content may increase the per-view income from Computing the normalizing constant of the corresponding likelihood requires the
customers who are willing to pay for the content. However, charging for content enumeration over all possible choice combinations. Thus, in large menus the
that was previously available free of charge might also lead to a consumer computation of the full likelihood becomes computationally prohibitive.
backlash and adversely affect brand value. To disentangle these potential effects, Historically, pseudo-likelihood approaches (e.g., Besag 1972, 1974) were used to
we exploit a quasi-experiment that recently occurred when the world’s largest sidestep this computational challenge. I show that the pseudo-likelihood is not
video streaming platform introduced a paywall (i.e., YouTube Red). Specifically, only inefficient but inconsistent when applied in a hierarchical setting, and how
we investigate how the paywall introduction affects (1) brand usage and algorithms that efficiently approximate ratios of normalizing constants result in
popularity (e.g., video views, channel subscribers), and (2) social media brand improved inference. Finally, I illustrate how the exchange algorithm (Moeller et
engagement (e.g., Twitter followers, Facebook likes, sentiment of user al., 2006; Murray et al., 2006) enables exact likelihood based inference when the
comments). To estimate these effects, we rely on a data set containing daily usage likelihood’s normalization constant is prohibitively expensive to compute.
and engagement metrics from various social media platforms for a period of more
than two years. The sample comprises three million videos, produced by 4200
human brands, and two billion textual user comments, analyzed using state-of-
the-art sentiment analysis techniques. The results suggest that the proportion of
negative comments increased directly after the paywall introduction, but this
effect wears out over time. Surprisingly, we find positive effects for brand usage
and popularity, suggesting a promotional effect of the paywall introduction.
73
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 74
74
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 75
n FD11 n FD12
Room 606, Alter Hall Room 745, Alter Hall
Emerging Markets Mobile Apps
Contributed Session Contributed Session
Chair: Matthew Osborne, University of Toronto, Toronto, ON, Canada, Chair: Yujie Deng, Clemson University, Central, SC,
[email protected], Avni M. Shah, Dilip Soman, [email protected], Chungsang Tom Lam
Jaclyn Lefkowitz, Andrew Fertig, Nina Mazar
1 - The Impact of Mobile Payment Service
1 - Product’s Characteristics as Drivers of Trickle-down and Reverse
Innovation: Evidence from the Food Industry Qiang Lu, University of Sydney, Discipline of Marketing, School of
Business - Econ Building H69, Sydney, 2006, Australia,
Verdiana Giannetti, PhD Candidate, Bocconi University, [email protected]
Via Roentgen 1, Milan, 20136, Italy,
[email protected], Gaia Rubera Although mobile payment has received considerable attention from companies,
little research has investigated how it might influence a company’s performance.
Nowadays, growth for multinationals comes largely from introducing innovations In this study, we attempt to shed light on this important topic by studying the
in emerging countries. However, this practice proves particularly challenging for impact of the introduction of a mobile payment service on consumer purchase
firms traditionally operating in developed countries. Adding further complexity, behavior. Based on existing theories of money and time, we construct a
such firms also have to defend their positions in developed countries from the rise conceptual framework and a set of testable hypotheses. To test the hypotheses, we
of emerging countries’ firms. Hence, it is now crucial to understand how to analyze a unique data set from a gasoline retail chain offering a mobile payment
accelerate the diffusion of innovations from emerging countries to developed ones service. Our results show that mobile payment can increase companies’ overall
(i.e., reverse innovation) and vice versa (i.e., trickle-down). A faster diffusion sales. We also find that mobile payment can activate the domain of money,
would in fact unlock additional sources of cash flow by disclosing new markets for thereby encouraging consumers who are price sensitive to choose pre-paid fuel
innovations. Using data on 127,782 food-products’ launches in 51 countries in card services. This impact is moderated by the time saved through the mobile
2001-2014, the paper answers four questions: (1) Which diffusion process occurs payment service. Furthermore, consumers who adopt a mobile payment service
faster: trickle-down or reverse innovation? (2) Which characteristics of tend to purchase less in each transaction; this is especially true for regular product
innovations accelerate trickle-down and reverse innovation? (3) Is there users. We also provide managerial implications.
heterogeneity in the effects of such characteristics across specific emerging
(developed) countries? (4) Does the speed of trickle-down (reverse innovation)
influence the performance of innovations? Results show that: reverse innovation
occurs faster than trickle-down; price, number of claims, and package size all play
75
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 76
76
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 77
77
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 78
78
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 79
n SA01
Room 32, Alter Hall n SA02
Online Advertising and Privacy Protection Room 33, Alter Hall
General Session Dynamic Pricing and Revenue Management:
Chair: Klaus Miller, Goethe University Frankfurt, Frankfurt am Main, Perishable and Seasonal Goods
60323, Germany, [email protected] General Session
Co-Chair: Bernd Skiera, Goethe University Frankfurt, Frankfurt am Chair: Mingyu Joo, Ohio State University, Columbus, OH, 43210,
Main, 60323, Germany, [email protected] United States, [email protected]
1 - Dynamic Selection Problems in Digital Advertising
Co-Chair: Robert E Sanders, The University of Chicago, Chicago, IL,
Navdeep S Sahni, Stanford University, 655 Knight Way,
United States, [email protected]
Knight Management Center, Stanford, CA, 94305, United States,
[email protected] 1 - An Empirical Test of Price Theories in the Market for
Seasonal Goods
Consumers may dynamically select themselves into seeing more, or fewer ads. A
Gonca Soysal, University of Arkansas, Sam M. Walton College of
consumer who saw an ad while browsing news on the internet might visit the
advertiser’s website and might, as a result, see more ads in the future. Dynamic Business, University of Arkansas, Fayetteville, AR, 72701,
selection of this nature makes it difficult to estimate the effects of frequency of United States, [email protected], Pradeep Chintagunta
advertising, even when the researcher conducts a randomized A/B experiment. Three theories have been proposed in the literature to describe the reason behind
This paper discusses the relevance of this problem, and provides a solution to sharp price declines observed over a product’s short lifecycle in seasonal
address it. (perishable) goods markets: Prices decline as a result of (1) decreasing opportunity
2 - The Impact of Privacy Policy on the Auction Market for Online costs over time; (2) firm’s uncertainty about relative product popularity at the
start of the season; (3) inter-temporal price discrimination. We first provide
Display Advertising empirically testable implications of each theory. Then, using price and demand
Garrett A. Johnson, Boston University, Boston, MA, United States, data from a large US specialty apparel retailer, we assess whether these theories
[email protected] have empirical support in the fashion apparel market. A unique feature of our
The advent of online advertising has simultaneously created unprecedented data that we leverage in our empirical analysis is the availability of inventory
opportunities for advertisers to target consumers and prompted privacy concerns information over time. Our results indicate that decreasing opportunity costs over
among consumers and regulators. This paper estimates the financial impact of time is the main driver behind the observed price declines. We do not find
privacy policies on the online display ad industry by applying an empirical model empirical support for the impact of the retailer’s initial uncertainty about relative
to a proprietary auction dataset. Two challenges complicate the analysis. First, product popularity or intertemporal price discrimination on observed price
while the advertisers are assumed to publicly observe tracking profiles, the declines in our data.
econometrician does not see this data. My model overcomes this challenge by 2 - The Effects of Menu Costs on Supply Chain Efficiency: Evidence
disentangling the unobserved premium paid for certain users from the observed
from Adoption of the Electronic Shelf Label Technology
bids. In order to simulate a market in which advertisers can no longer track users,
I set the unobserved bid premium’s variance to zero. Second, the data provider Ioannis Stamatopoulos, University of Texas at Austin, McCombs
uses a novel auction mechanism in which first-price bidders and second-price School of Business, 2110 Speedway B6000, Austin, TX, 78705,
bidders operate concurrently. I develop new techniques to analyze these hybrid United States, [email protected],
auctions. I consider three privacy policies that vary by the degree of user choice. Achal Bassamboo, Antonio Moreno
My results suggest that online publisher revenues drop by 3.9% under an opt-out
We use the adoption of electronic shelf labels (ESLs) by a major international
policy, 34.6% under an opt-in policy, and 38.5% under a tracking ban. Total
grocery retailer in 2015 in the United Kingdom to identify the effects of reducing
advertiser surplus drops by 4.6%, 4’39%, and 45.5% respectively.
physical menu costs (operational costs of price adjustment) on supply chain
3 - Digital Identity, Privacy and the Inequality of Outcomes: efficiency. The ESL technology essentially eliminates the physical costs associated
The Case of the Aadhaar Initiative with price adjustment (e.g., costs of printing and distributing price tags). We find
Anuj Kapoor, University of Utah, David Eccles School of Business, that the elimination of physical menu costs benefits all supply chain stakeholders
(retailer, consumers, suppliers). In our setting, daily revenues increased, the
1655 Campus Center Drive, Salt Lake City, Utah, UT, 84102, average price per unit sold decreased, and daily sales volumes increased as a result
United States, [email protected], Avi Goldfarb, of ESLs. We also find that ESL adoption increased price-adjustment volume,
Catherine E. Tucker decreased the average size of a price adjustment, and decreased the batching of
The question of digital identity poses many challenges for both policymakers and price changes across different products. Finally, we find that ESL adoption had a
firms. On the one hand digital identity poses unique privacy risks to the statistically significant effect on the volume of downward price changes, but not
individual. On the other hand, without a digital identity people may be excluded on the volume of upward price changes, which explains the direction of the
from digital services. Therefore, to prevent digital exclusion it is natural for change in operational outcomes.
policymakers to try and facilitate digital identities for the poor. To study the 3 - Pre-ordering And Manufacturer’s Return Policies for Durable
tensions inherent in this digitization process we study the digital transition to
Goods: Theory and Empirical Evidence
Aadhaar - which is a unique digital identifier the Indian government has
sponsored for its citizens. We document a unique distortion in this process. We Jong Yeob Kim, PhD Student, NYU Stern School of Business,
show that poor people are more likely to experience issues with establishing a Tisch Hall 921, 40 West Fourth Street, New York, NY, 10012,
digital identity due to transcription errors in their details. To identify the causal United States, [email protected], Masakazu Ishihara
effects of these digital errors we exploit variation in the quality of workers This paper theoretically examines the roles of manufacturers’ return policies in
assigned to each region to facilitate this digital transformation process. We durable goods markets when forward-looking consumers, retailers, and
evaluate whether these transcription errors matter we then turn to whether study manufacturer face uncertainty about product quality and consumers have an
whether these delays affect real financial outcomes, and find evidence that they option to pre-order. Using a simple multi-period vertical differentiation model, we
do. investigate the impact of manufacturers’ return policies on equilibrium purchase
4 - Economic Damage of Cookie Lifetime Restrictions decisions, retailer’s pricing and inventory stocking decisions, and manufacturer’s
Klaus Miller, Goethe-University, Theodor-W.-Adorno-Platz 4, wholesale pricing decisions. Our goal is to understand the conditions under which
it is optimal for a manufacturer to accept returns from retailers. We then test our
Frankfurt am Main, 60323, Germany, theoretical predictions using data from the U.S. and Japanese video game
[email protected], Bernd Skiera markets. In the U.S. market, video game publishers accept returns from retailers,
Over the past few years, regulators have begun to consider restricting a cookie’s but in the Japanese market, publishers do not. The novel aspects of our data are
lifetime or even banning cookies altogether as a way to protect consumer privacy. that we observe pre-order sales and retailer’s inventory stocking decisions. We
Most of this debate has taken place in the absence of any quantified cost-benefit show that our theoretical predictions are empirically supported in these markets.
analysis. To begin to fill this gap in the discourse, we estimate the potential
economic damage of lifespan restrictions on cookies. Our analysis is based on an
empirical study on cookies of 54,127 users who received about 130 million ad
impressions over 2.5 years. Only 22% of all cookies increase their daily value
over time but the value of that quantile represent 61% of the value of all cookies.
This analysis suggests that restricting their lifetime to one year as the European
79
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 80
81
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 82
82
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 83
n SA09 n SA10
Room 239, Alter Hall Room 605, Alter Hall
Product Choice & Assortment Demand, Choice and Behavior
Contributed Session Contributed Session
Chair: Roozbeh Iranikermani, Penn State University, 920 Oakwood Chair: Rudolf-Harri Oberg, Duke University, Durham, NC,
Avenue, State College, PA, 16803, United States, [email protected] [email protected]
1 - Product Line Design with Superior Information on Consumers’ 1 - Store, Category, and Brand Choice
Preferences: Implications of Data Aggregation Rudolf-Harri Oberg, Phd Candidate, Duke University, Durham,
Zibin Xu, Shanghai Jiao Tong University, Shanghai, 210010, NC, United States, [email protected], Andres I Musalem, Carl F Mela
China, [email protected], Anthony Dukes In many retail settings, understanding consumer demand is the basis for the study
Consumers’ initial perceptions for new products are often noisy indicators of their of firm strategy and public policy. In this paper, we develop a direct utility
intrinsic valuations. If the noises in consumers’ perceptions are correlated, then a framework to model consumer demand for brands, categories, and stores. While
firm can acquire superior information over consumers by collecting aggregate these decisions have been considered separately in the literature, we show that a
marketing research data (e.g. surveys or conjoint studies). But is superior unified treatment of store, category, and brand demand captures novel
information ever unprofitable? Are consumers better off when the firm obtains substitution patterns and provides a much richer picture in many applications. We
superior information? Do they receive better fitting products or simply have more estimate the model on household panel data and show that it can be used to
surplus extracted? In a monopoly model of product line design with imperfectly study in-store marketing and health policy issues. For in-store marketing, we
informed consumers, we find that consumers’ rational suspicions may prevent the identify to what extent supermarkets in our sample engage in loss-leader pricing,
firm from fully exploiting its superior information. In addition, allowing the firm i.e., we identify brands which have the highest incremental effect from other
to collect consumer data may be strictly Pareto improving. stores and low margins. For health policy, we study the effectiveness of various
policy measures (health tax on sugar items, subsidy for healthy goods, food
2 - Heterogeneous Veblen Effects: Implications for stamps) in terms of their impact on nutrition outcomes and consumer welfare.
Assortment Decisions
Rafael Becerril Arreola, USC Darla Moore School of Business, 1014 2 - The Estimation of Discount Factors using Field Experiments
Greene Street, Moore School of Business, Columbia, SC, 29208, Meng Li, Rutgers University, 227 Penn Street, Camden, NJ, 08102,
United States, [email protected] United States, [email protected], Arun Gopalakrishnan,
Raghuram Iyengar
Veblen effects occur when product prices serve as signals of consumer wealth and
thus influence the utility that products offer to consumers. This study quantifies Determining how far ahead consumers plan their consumption is important for
heterogeneous Veblen effects with a quasi-experimental approach that addresses many marketing applications. Many contexts, however, present limitations in the
identification threats relevant to social influence and supply factors while estimation of discount factors due to insufficient variation in the observed data.
controlling for alternative social processes and the effects of prices on affordability As an example, consider a customer making data usage decisions over a monthly
and perceived quality. Using a four-dimensional panel data on car rentals, the billing cycle. Given the nonlinear pricing of data, usage decisions that the
study explains rental choices in terms of the different levels of relative wealth customer makes early on in the month have implications for her subsequent
signaled by different vehicles at different locations. The individual-level estimates usage during the month. Whether the customer is a light or heavy user of data,
reveal that both wealth signaling and counter-signaling are common among however, can change the extent to which their time discounting preferences can
travelers and that the average product-specific elasticity of choice probability with be recovered from observed decisions. In this paper, we use the context of
respect to signaled wealth is -0.65, about a third as large as the elasticity with cellphone data consumption and combine theory with a carefully designed field
respect to rental fees. A simulation study shows that profits can increase by at experiment to showcase the challenges in identifying the discount factor. The field
least 1% if the rental assortment is optimized to cater heterogeneous preferences experiment manipulates the amount of data available to customers and thereby
for wealth signaling. leads to variation in the state space for subsets of customers whose discounting
parameters would otherwise be poorly recovered. We show that both the utility
3 - Variety Seeking, a Generalized State Dependent Variable Based function and discounting primitives can be empirically estimated for a broader
on Product Attribute Space range of consumers using the field experiment. The estimation helps improve
Roozbeh Iranikermani, PhD Candidate, Penn State University, managerial insights on the willingness of these consumers to pay for data.
920 Oakwood Avenue, State College, PA, 16803, United States,
[email protected], Edward Jaenicke
Consumers’ tendency to substitute different types of the same product in their n SA11
baskets or search for ideal bliss point via diversification is generally regarded as
variety-seeking behavior, and researchers most often operationalize this concept Room 606, Alter Hall
by observing product-switching behavior from one purchase occasion to another. Experimental Research: Choice & Information
What all variety-seeking measurement models have in common is that they
operationalize this behavior by observing shopping patterns at the brand level. Contributed Session
Thus, variety seeking is the negation of brand loyalty. In this paper, we propose to
Chair: Siddhartha Sharma, Carnegie Mellon University, Pittsburgh, PA,
generalize the operational concept of variety seeking by focusing on the
differences of among the product attributes that underlie the brand differences. [email protected]
We construct a variety-seeking index that measures variety via relative Euclidian 1 - Prelaunch Demand Estimation
distances in attribute space. Because a brand identifier could therefore be just one
of several product attributes, a more traditional definition of variety seeking based Xinyu Cao, MIT Sloan School of Management, Cambridge, MA,
only on brands would be a special case of this new, more general, variety-seeking United States, [email protected]
index. We use the Nielsen Homescan dataset for the U.S. beer retail market and Demand estimation is important for new-product strategies, but is challenging in
calculate this inex during the years 2009-2012. Our results show that, adding this the absence of actual sales data. We develop a cost-effective method to estimate
index to the brand choice model, enhances more explanatory power to the model the demand of new products based on incentive-aligned choice experiments. Our
and improves estimation and analysis of consumer preferences. Now, it is possible premise is that there exists a structural relationship between manifested demand
to capture brand-loyalty and variety-seeking at the same time in the same and the probability of consumer choice being realized. We illustrate the
household independently. Further, the model can be extended to other situations mechanism using a theory model, in which consumers learn their product
that require similar state-dependent variables. This could facilitate polling by valuation through costly effort and their effort incentive depends on the
political parties, or simplifying risk assessment for insurance companies, banks, realization probability. We run a large-scale field experiment on a mobile game
and lenders. platform, where we randomize the price and realization probability when selling
a new product. The data support our theoretical prediction and the decision effort
mechanism. We then estimate a structural model of consumer choice. The
structural estimates allow us to infer actual demand with reasonable accuracy
using inexpensive choice experiments with small to moderate realization
probabilities.
83
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 84
84
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 85
85
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 86
n SA15 Research has stated that the use of a credit card as a payment mechanism
increases the propensity to spend as compared to cash in otherwise identical
Room 603, Alter Hall purchase situations (Feinberg 1986; Hirschman 1979; Prelec and Simester 2001;
Soman 2001), referred to as the credit card premium. Though there is research
CB – Information Processing & Adoption conducted on the effects of credit card usage on consumers’ purchasing behaviour
Contributed Session and perception of product features in marketing literature, less work is done to
understand the adoption of payment methods various consumer demographics on
Chair: Xuefeng Liu, Loyola University Maryland, 4501 N Charles Street, use of different payment methods. Recent research on emerging markets also
Baltimore, MD, 21210, United States, [email protected] highlighted the difference in adoption of modern retail across socio-economic
classes in India. (Narayan, Rao and Sudhir, 2015). We extend the literature on
1 - Impact of Friendships Among Customers on their Perceived Value
differences in consumer adoption across socio economic classes by studying
from Consumption differences in the use and adoption of alternative payment systems in the
Prashant Mishra, Professor, Indian Institute of Management emerging markets. Through a unique natural experiment, we are able to assess
Calcutta, D.H.Road, Joka, Kolkata, 700104, India, the impact of alternative payment methods on revenues of two retailing modes-
[email protected], Diptiman Banerji, Ramendra Singh, home delivery and in store retailing for a restaurant in India. In an
Rebecca G. Adams, Rajiv Kumar unprecedented government action, high value currency denominations were
suddenly withdrawn, leading to cash scarcity and income impact, which differed
Value creation is a key task for marketers. Researchers suggest that value is across socio economic classes. A unique real data set allows us to study impact of
created by customers in their usage experiences—a notion of value referred to as scarcity on consumption habits and differences in approach to alternate payment
“value-in-use” or “customer perceived value” (CPV)—and each interaction at a methods across socio economic classes across delivery formats. We use a
touchpoint influences such value creation. Our study investigates the effect of difference in difference approach to understand the difference across socio
consumer friendships on CPV in the context of joint consumption. We economic classes and delivery formats. We extend generalizability of the findings
hypothesize that the presence of a friend increases the CPV from a joint by conducting a related laboratory experiment to address generalizability
consumption experience (assuming a positive interaction). We suggest that an concerns. Our study contributes to the literature exploring understanding of
increase in positive affect (PA) and thought confidence (TC) levels mediate this adoption formats across economic classes in emerging markets. The results of the
relationship. We also argue that the change in CPV leads to a corresponding study would help marketers understand the adoption of alternate payment
change in re-patronage intention (RPI). We further posit that mental intangibility methods across socio economic classes and hence help in pricing and payment
(MI) acts as a moderating variable for the path from PA to CPV and from TC to method decisions.
CPV. We test our hypotheses over three experimental studies. The first two are 2
X 2 experiments consisting of two conditions (friendship: alone and with friend) 4 - The Interactive Impact of Headline and Body Copy on
X two service situations (MI: low and high) and a positive interaction is primed. Ad Effectiveness
In both these studies, we get support for the mediating effect of PA and TC (with Xuefeng Liu, Assistant Professor, Loyola University Maryland,
the latter being stronger), and the moderating effect of MI. Our third experiment 4501 N Charles Street, Baltimore, MD, 21210, United States,
(a simple two-way experiment, with no manipulation of MI) tests a potential
[email protected], Jason Zhang, Jibo He
boundary condition of the hypothesized friend’s influence when there is a
lowering of PA. We find that consistent with the lowering of PA, CPV significantly In both online and offline settings, it is common for an ad to include a headline
reduces in the presence of a friend as compared to the alone condition.Overall, and body copy. The headline presents a positioning claim that highlights key
we get empirical support that the presence of a friend affects CPV, and this change selling points, whereas the body copy presents additional information. In
is mediated by PA and TC, with TC emerging as a stronger mediator. Our examining the impact of the ad, prior research treats the effect of ad headline and
experiments also support the notion of MI as a moderator. Further, we show that body copy as additive. This research, however, shows the effect of headline and
the presence of a friend influences RPI as well. We also study the boundary body copy as interactive, in that the different positioning strategies of a headline
conditions of the impact of friendship, and the results strengthen our conclusions. (e.g., focus on specific attributes vs. abstract benefits) would bias how consumers
In all, over three experimental studies, we find considerable support for our process the information in the body copy and thus their product attitudes.
hypotheses. Our findings are significant from both a theoretical and a managerial Following the hypothesis-testing theory (Deighton 1984), we propose that
perspective. consumers view positioning claims in ad headlines as tentative hypotheses and
use the information in the accompanying body copy to test these hypotheses.
Consistent with this perspective, we predict that relative to abstract
positioning headlines, attribute-specific headlines lead to more (less) favorable
product attitudes when data from the body copy confirm (do not confirm)
positioning claims. Several lab experiments support these predictions, including
an eye-tracking study offering evidence on the proposed processes. Our research
extends the hypothesis-testing theory by demonstrating that tentative hypotheses
can be formed and tested within an ad (vs. tested with information outside the
ad, as in the original theory). Additionally, we show that headlines influence
consumers’ processing of body copy, a mechanism largely ignored by previous
research—offering practical guidelines for the development of effective headlines
and body copy.
86
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 87
87
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 88
88
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 89
n SB05
n SB04 Room 232, Alter Hall
Room 35, Alter Hall Mobile, Algorithm, and Artificial Intelligence (AI)
Digital Transformation of the Marketing Paradigm Session VII: Mobile Usage and Customer Welfare
General Session General Session
Chair: Andreas Lanz, University of Mannheim, C8, 20, Mannheim, Chair: Xueming Luo, Temple University, Philadelphia, PA, 19122,
68159, Germany, [email protected] United States, [email protected]
1 - Putting Brands in Context Co-Chair: Yuchi Zhang, Temple University, Philadelphia, PA, 19122,
Verena Schoenmueller, Columbia University, 930 St Nicholas United States, [email protected]
Avenue, Apartment 6, New York, NY, 10032, United States,
1 - Do Spoilers Spoil? An Empirical Study of Movie Reviews using
[email protected], Oded Netzer, Florian Stahl
Topic Modeling
One of the most useful bases for market segmentation is preferences-based Jun Hyun (Joseph) Ryoo, [email protected]
segmentation. However, one of the difficulties in segmenting consumers based on
their preferences has been the limited information regarding consumers’ brand Most movie review sites allow users to post spoilers, which reveal a certain level
preferences especially across product categories. Social media platforms such as of movie plots. Conventional wisdom suggests that giving away the plot is likely
Facebook and Twitter provide readily available insights regarding the brand to reduce the enjoyment of movie consumption of potential moviegoers and
preferences of millions of consumers in the form of brand likes and followerships. therefore hurts the box office revenue. However, spoilers might also signal the
Building on a large Twitter dataset incorporating information regarding brand credibility of review contents and therefore enhance the informative or
followerships across more than 100M followers of more than 600 brands this persuasive effects of online word-of-mouth (WOM). In this paper, we investigate
paper aims to derive “consumer brand personas”, i.e., consumer segments based the effect of spoilers on aggregate box office sales based on a dataset of movies
on their brand affinity across product categories. To derive these latent released in the United States from 2013 to 2015. We employ a correlated topic
dimensions of consumer preferences, we build on matrix factorization techniques modeling method to extract and quantify the plot element of all spoiler reviews
that allow us to handle large amount of sparse data and let us uncover different on IMDb.com. We find that the amount of spoiler information has a net positive
latent brand personas based on the brands a Twitter user follows. effect on box office revenue. This positive effect is more pronounced for lower
rated movies, the users’ consumption of which tend to rely more on movie
2 - The Role of Incentivized Reviews: a Dynamic Perspective reviews. We also find a positive interaction effect between the level of plot-related
Cindy Zhao, University of Maryland, College Park, MD, 20742, contents in spoilers and review volume. These patterns are consistent with the
United States, [email protected], Michael Trusov credibility-enhancement effect of spoilers. Furthermore, controlling for the pre-
consumption spoilers, we find that the plot-related information in
Incentivized campaigns have become an effective and efficient way to stimulate post-consumption spoilers also leads to more theatrical visits, suggesting that the
electronic word-of mouth (eWOM) in terms of both volume and valence. consumption of movies and spoiler reviews partake in a complementary rather
However, it also has the potential to negatively affect eWOM in the long run. In than a substitutive relationship.
this paper, we investigate the impact of incentivized reviews on subsequent
reviewing behavior. We empirically model the dynamic effect of incentivized 2 - User Contribution and its Social-Welfare Value in a Mobile App for
reviews using Amazon.com review data collected across multiple product Real-Time Traffic Information Around Urban Areas
categories. Our results show that after the incentivized review enters the system, Chenhui Guo, Michigan State University, 632 Bogue Street, Room
ratings of subsequent organic reviews first decrease but then recover over time. N260, East Lansing, MI, 48824, United States, [email protected]
The temporary decline in ratings is significant after we capture the time and order
trend as well as other factors known to influence product ratings. We employ We investigate the effectiveness of mobile user contribution with a GPS
natural language processing techniques to investigate how the composition of navigation app, Waze. Waze app users contribute by generating real-time traffic
subsequent reviews changes due to the incentive disclosure. content: primary contribution by posting alerts as well as additional contribution,
feedback by commenting and confirmation by putting thumbs-up on the alerts.
3 - Board Interlock Networks and Marketing Department Power Our findings from a spatial data approach explain how user contribution mitigates
Peter Ebbes, HEC Paris, Paris, France, [email protected], traffic congestion around New York City: reductions of the number, duration, and
Frank Germann, Raj Grewal length of traffic jams. Regarding the primary contribution, while the first alerts—if
at least a single alert is initially posted around traffic jams—are effective in
It is now well established that marketing department power in firms is important mitigating traffic jams, follow-up alerts added over time—a high degree of the
for firm outcomes including firm performance. Surprisingly, however, marketing number of total alerts—are not. This implies that not the amount of total alerts
department power in many firms is low, and some research even suggests that it but a few initial alerts may improve driving conditions: while the few initial alerts
is decreasing. Seeking to address this apparent disconnect, the authors propose generate new and useful information, the follow-up alerts do not add meaningful
that the board of directors is a critical but neglected driver of marketing content due to information redundancy and overload. Additionally, increased
department power in firms. In particular, they examine how directors’ marketing comments (feedback) and thumbs-up (confirmation) both reduce the number of
exposure through board service at other firms affects marketing department traffic jams whereas they have different effects on the duration and length of
power in focal firms. Considering a large sample spanning the 2006 - 2013 period, traffic jams. Confirmation reduces the duration and length of traffic jams whereas
they find that marketing department power in board-interlocked firms feedback does not. This shows that the effectiveness of user contribution is
significantly and positively drives marketing department power in focal firms. differentiated by the different aspects of traffic congestion: its numeric volume,
Moreover, consistent with an information sharing interpretation, the magnitude temporal duration, and spatial length. We discuss the effectiveness of mobile user
of this effect varies based on the focal firm’s board-interlock network position, contribution, crowdsourced by Waze app users, and practical implications on its
where it strengthens as the focal firm’s boardinterlock centrality increases, but value for public welfare with the driving quality around urban areas.
89
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 90
90
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 91
91
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 92
92
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 93
93
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 94
95
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 96
96
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 97
97
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 98
98
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 99
n SC08 n SC09
Room 238, Alter Hall Room 239, Alter Hall
Consumer Decision-Making – Signaling & Influences Promotions & Reward Programs
Contributed Session Contributed Session
Chair: Yingge Qu, Mississippi State University, 2212 5th Street, #410, Chair: Anthony Koschmann, Eastern Michigan University, 7415 Willow
Meridian, MS, 39301, United States, [email protected] Creek Dr, Ypsilanti, MI, 48197, United States, [email protected]
1 - Time to Next Member’s Travel Booking: A Shared Frailty 1 - Premium Private Label Sourcing from a Dual Brander:
Approach to Account for Member’s Specific Unobserved Factors Competition and Negotiation
Mouna Sebri, Assistant professor, Sherbrooke university, 2500 S. Chan Choi, Professor, Rutgers Business School, 100 Rockafeller
Boulevard de l’Université - Immeuble K1, Sherbrooke, QC, J1K Road, Piscataway, NJ, 08854, United States, [email protected]
2R1, Canada, [email protected] As private labels become proliferated, more retailers are introducing premium PLs
Within a coalition loyalty program where an airline company is a major source of that oftentimes replace marginal national brands. A premium PL can be sourced
members’ accumulation as well as an attractive choice in terms of rewards, either from the NB manufacturer who wants to utilize excess capacity or from a
tracking every travel opportunity from the member perspective is crucial for both dedicated PL manufacturer who can produce a high-quality imitation. It is natural
the airline company and the loyalty program sustainability. Determining flight- to assume that the PL from the former source is identical to the corresponding NB
booking pattern in the travel industry is crucial to divert flight opportunities from except for the branding and packaging. In this paper, we examine a retailer’s
competition and support the airline share of wallet in the loyalty program. The problem of tiered PL sourcing, in which a premium PL is supplied by the NB
key challenge of such non-contractual setting, is how to differentiate those manufacturer (dual brander), and an economy PL is supplied by a dedicated PL
travelers who have ended their relationship with the airline company without supplier. We decompose the value of a product into three components: the NB’s
informing them and without necessary expiring from the loyalty program from brand equity, the retailer’s reputation, and the intrinsic quality of the NB. In this
those who are simply in the midst of a long hiatus between travels. We develop a distribution channel, the NB’s wholesale and retail prices are determined by the
model to predict future traveling patterns for a customer base that can be traditional bilateral Nash game. However, the premium PL’s transfer price is
described by these structural characteristics. A shared frailty model that accounts determined through a profit-sharing negotiation between the channel members.
for member’s specific unobserved factors is proposed to accurately predict the From an equilibrium-negotiation solution, we derive profit implications of each of
time to next travel with the airline company at the member level. Travel bookings the value components as well as the negotiation power of the retailer.
at various future time windows are predicted to better adapt campaigns objectives Interestingly, even if the retailer holds a strong negotiation power, we find it
and allow sufficient lead-time for marketing actions. optimal for the retailer to leave some chips on the table for the NB manufacturer.
2 - The Impact of Future Event Markers on Intertemporal Decision 2 - Optimal Pricing of Points in Points Plus Cash Reward Programs
Among Different Decision Makers Ricardo Montoya, University of Chile, Republica 701, Santiago,
Jessy Zhang, Tongji University, Shanghai, China, 8370439, Chile, [email protected]
[email protected] Customers in Reward Programs (RPs) typically accumulate points for their
Intertemporal decision is a normal decision-making process happened in purchases that can be redeemed later for rewards. Recently, some RPs offer the
consumers’ daily lives. It’s about a two-dimensional choice between conserving option of combining a customer’s points with cash to redeem them for products in
waiting time and harvesting bigger future profit. Although many scholars have a Points Plus Cash (PPC) frame. The cash in this scheme helps to buy the needed
studied the influential factors in intertemporal decision process from the points to complete the full-point equivalent needed to redeem for the reward.
perspective of decision-making framework, the impact of planned future event The price of these points need to be determined by the company considering
markers on intertemporal decision-making is seldom studied, which can be various factors such as the consumers’ consumption heterogeneity and
exactly observed from real life. This paper tries to explore the different effects of willingness to pay for the points, the value of the product, and the cost of that
event markers on intertemporal decision across different time-perspective product. The goal of this research is to determine the optimal price in a PPC
decision makers using experimental method: 1.For present-oriented decision scheme taking these factors into account. The proposed framework considers
makers, more events between a small-sooner and larger-later option will shorten differences in consumption rates and product valuations by customers that allow
their time perceptions and thus increase possibility choosing larger-later; 2.For a profit-maximizing firm to determine the optimal dynamic price. Our main result
future-oriented decision makers, more events between a small-sooner and larger- characterizes the optimal price as a function of the time remaining to the end of
later option will lengthen their time perceptions and thus increase possibility the selling horizon. We show that the price increases in time and decreases
choosing small-sooner. Findings in this study help to further understand regarding the number of points required to be redeemed for the product. Demand
intertemporal decision-making process and suggest consumers to make rational at the optimal dynamic price also increases over time. The optimal pricing
decisions by adaptive planning. uncovers three types of customers. A first segment prefers buying at the
beginning of the selling horizon, a second group prefers buying at the end of the
3 - Modeling Consumer’s Contractual Decision in a Continuous selling horizon, and a third group buys during the selling horizon. We use the
Innovation B2b Market with a Forward-looking Dynamic Approach developed framework to compare the proposed pricing strategy to other pricing
Yingge Qu, Assistant Professor, Mississippi State University, options such as constant pricing and also to explore extensions to our basic
2212 5th Street, # 410, Meridian, MS, 39301, United States, model. The analysis shows that the main results are fairly robust to different
[email protected], V. Kumar, Yi Zhao assumptions.
The technology service of cloud computing in the B2B market possesses several 3 - Consumer Response to Price and Promotions Before and After
features that challenge the service provider’s strategic decision making. First, B2B the Great Recession
buyers sign a contract when purchasing the service. Second, the firm’s pricing Anthony Koschmann, Eastern Michigan University,
strategies of the service ties to the length of the service contract that the 7415 Willow Creek Dr, Ypsilanti, MI, 48197, United States,
consumer signs. Third, to maintain advantages in the market, the service provider [email protected]
needs to keep on improving the service-related technology. Last, the B2B buyer’s
decisions are more sophisticated and strategic when balancing the benefits and Previous research into the effects of recessions on consumer purchases have
the risks of the service contract. To address service provider’s concerns, we typically emphasized price sensitivity. Yet, price remains just one of the tools that
develop a structural modeling approach to address the following research firms use to encourage consumption during and after a recession. This study
questions: 1) how to quantitatively evaluate the B2B consumer’s contractual examines consumer response to types of promotions, such as price reductions
decision process in the continuously developing technology service market? 2) alone or price reductions in conjunction with another promotional device
when the service provider adjusts the technology adoption in the service, how (manufacturer coupons, weekly feature advertising, and in-store displays)
will the B2B buyer’s contractual decisions change? and 3) what is the optimal following the Great Recession. The severity of the Great Recession, from both its
pricing strategy for the service provider to maximize the contract revenue? Our depth and recovery, suggests consumers increased their purchase response levels
study has both modeling and managerial contributions. From modeling and exhibited an increasing rate of response to firms’ price and promotions well
perspective, we contribute to the forward-looking dynamic models by integrating after the Great Recession had passed (in this case, six years into the recovery).
three key features of the market, e.g. the B2B buyer’s dynamic contractual Using annualized U.S. consumer panel purchases from 1998-2015 for 581
decisions, the technology evolution and the firm’s pricing strategy, into one product categories, the results show that the Great Recession of 2008-2009
holistic modeling framework. More importantly, these key features are not altered the promotional trajectories of some product categories, but not
directly used as control variables or covariates in the utility function but universally. Different promotional types exhibited differential effects: the level of
integrated into the dynamic programming process as state variables determining price reduction and manufacturer coupon response increased, but feature
the contract-associated value. Our study also provides substantive insights. We advertising and in-store display response decreased among more product
address that, service providers need to realize that the technology evolution can categories than increased. Furthermore, the rate of response decreased for nearly
be both beneficial and risky to the service provider. By counterfactual analysis, we three-fourths of product categories for price reductions following the Great
identify an optimal pricing strategy to help the service provider optimizing the Recession; the interpretation is that consumer responsiveness to price reductions
contract revenue. ramped up prior to the Great Recession, and rate of response plateaued (or even
decreased) following the Great Recession.
99
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 100
n SC10
that time frame. Using a difference-in-difference methodology, for each artist who
died, we compare the daily physical (more material purchase) and digital (more
experiential purchase) album sales a week before death and a week after death to
Room 605, Alter Hall any living artist’s physical and digital album sales over the same period. We find
Digital Entertainment empirical support for our hypotheses - artists who died suddenly (e.g. accident,
suicide, etc.), highlighting the ‘uncertainty’ dimension of mortality, experienced a
Contributed Session significant increase in physical album sales relative to digital album sales; artists
Chair: Stan Renard, The University of Texas at San Antonio, One UTSA whose death was less sudden, more expected with an increase in time, (e.g. died
Circle, San Antonio, TX, 78249, United States, [email protected] from terminal illness, old age, etc.), highlighting the ‘time scarcity’ dimension,
experienced a significant increase in digital album sales relative to physical album
1 - How Playlists Shape Artist Success on Spotify sales. We discuss managerial implications beyond the music industry.
Sean N. Bruggemann, Swiss Federal Institute of Technology,
Zurich, Switzerland, [email protected], Anand V. Bodapati, 4 - Beyond the Grave: Variables Contributing to Increased Music
Randolph E. Bucklin Consumption After an Artist’s Death
Stan Renard, Assistant Professor of Music Marketing,
How critical are playlists in determining artist success on music streaming The University of Texas at San Antonio, One UTSA Circle,
platforms such as Spotify? On-demand streams for music have grown by 76.4%
San Antonio, TX, 78249, United States, [email protected],
in 2016, reaching an all-time high of 252 billion streams (Nielsen, 2017).
Streaming services such as Spotify or Apple Music count 112 million paying Richard T. Gretz
customers and contribute to about a third of music labels’ total revenues. The recent passing of notable recording artists such as Prince, David Bowie, Tom
Listening to music via Spotify differs from conventional modes of music Petty, and Chuck Berry has generated a surge in music sales associated with those
consumption in that users rely heavily on playlists - i.e. self -or third-party artists. However, the impact of the death of these artists on the sales of other
curated collections of tracks: 50% of paying subscribers to Spotify use playlists on musicians is not well understood. We use a difference-in-difference approach to
every usage occasion and 90% of paying users have already listened to playlists at assess the impact of an artist’s death on 1) the sales of the artist’s albums and
least once (Music Watch, 2016). Playlists stimulate demand and popularity for songs, 2) the sales of living artists’ albums and songs in the same genre as the
those artists who are featured on them and can be a critical marketing tool for artist who died, and 3) the sales of living artists’ albums and songs in a different
popularizing artists’ musical products and brands. We quantify the effect of genre as the artist who died. We leverage a unique dataset where we observe
playlists on artist popularity by analyzing a weekly data set of tracks that were daily song and album sales (both physical and digital) for all artists who were on
added to or removed from Spotify playlists with more than 500,000 followers, in the Billboard Top 200 and/or passed away from January 1, 2015 through
the period ranging from July 2017 through January 2018. Our data show the December 31, 2017. For each artist who died, we compare that artist’s song and
track’s position within the playlist and artist popularity, which is a measure of the album sales several days before death and several days after to all living artists on
number of times the artist has been streamed relative to all other artists. Findings the market during that same period. We are able to identify what characteristics
from a disaggregated model with artist-week as the unit of analysis hint at both are associated with an increase or decrease in album and song sales due to a
contemporaneous and lagged effects of playlists on artist popularity. The effects fellow artist’s death. Our research has immediate applicability to the music
hold after controlling for artist heterogeneity through artist-level fixed effects and industry, however our results also apply more broadly to product portfolio
changes to the artist’s follower base. We additionally control for artist-specific management. Our distinctive setting allows us to obtain insight into the
changes to their popularity on an unrelated streaming platform (i.e. views on performance of other members of a product portfolio when there is an exogenous
YouTube). (and catastrophic) shock to one member.
2 - A Double-edged Offer: Gamers’ Opposite Responds to
Game Updates
Jihyeon Hyeong, Yonsei University, Seoul, Korea, Republic of, n SC11
[email protected], Kangjun Choi, Sujin Ko, Jaeyoung Lee,
Tae-Hyung Pyo Room 606, Alter Hall
This research aims to expand our understanding of game updates and offer Healthcare-Services
strategic plans for the game industry. In the game industry, game updates have Contributed Session
long been recognized as a classical way of motivating players with the
introduction of new game contents and rules. Despite this, scant research has Chair: Lu Liu, Syracuse University, 721 University Avenue, Syracuse,
examined whether updates substantively satisfy the nuts and bolts required for NY, 13244-2450, United States, [email protected]
marketing actions— to see whether updates can positively influence gameplay. 1 - Insurer Competition and Premium Dispersion on
This idea, then, leaves the question of whether game updates are desirable even
Healthcare Marketplaces
to those players who are eager to consolidate their status in competition. It seems
counterintuitive given that competitive games, i.e. multiplayer games, usually Ming Lei, Purdue University, 403 W State Street, West Lafayette,
take advantages from competition among players, unlike noncompetitive games, IN, 47907, United States, [email protected], Qiang Liu, Ting Zhu
i.e. single player games. However, for players motivated by competition, we Affordable Care Act (ACA) introduced health insurance marketplaces that
expect that there might exist an unintended and even the exact reverse effect of provide ACA qualified health insurance products of various attributes with high
updates; players in high status would be discouraged by a new competition rule, subsidies as well as heavy regulations. This paper studies the impact of insurer
as it potentially threatens their dominance, and therefore may end up leaving the entry on insurance premiums in ACA marketplaces. We utilize the public data
game. For the empirical analysis of our research question, we collect the from federal run marketplaces and construct a panel of insurance premium and
individual play data from a game platform site and additional information on characteristics for all available plans in every county of 34 states. Our initial
games from the official game websites. Through developing a hazards model, we finding is consistent with previous literature in that insurer entry would lower
compare the probability of the individual playing the game after updates between premium for a specific type of plan. In addition, we find the premium dispersion
the two types of games and investigate the interaction effect of the players’ status between high quality and low quality plans increases with competition. We
and the game types on this probability. Our results show that updates increase the further explore possible mechanisms of this dispersion and discuss managerial and
probability of individuals playing multiplayer games less than single player games. public policy implications of our findings.
Furthermore, these results are derived from the trade-off between the opposite
behavior of players in high- and low status in multiplayer games: players in high 2 - The Effect of General Health Check on Healthcare Utilization:
status are less likely to play the game after the update, while players in low status Tackling the Self-selection Bias
are more likely to play the game after the update. We expect that our findings Sung Wook Yoon, KAIST, Seoul, Korea, Republic of,
provide new insights into the deliberate strategy on how to attract and retain [email protected], Duk Bin Jun, Sungho Park
players using updates.
The general health check is one of the most common preventive health care
3 - Do Musicians Sell More Physical Albums or Digital Downloads measures in many countries. This typically involves the visiting of general
After They Die? Uncertainty vs. Time Scarcity in Mortality Salience practitioners or healthcare institutions to detect disease and/or risk factors for
Bingxuan Guo, PhD Student, University of Texas at San Antonio, disease by healthy adults who do not feel ill. Researchers have examined the
5803 USTA Blvd, Apt 3308, San Antonio, TX, 78249, United effect of the general health check on various health outcomes and subsequent
States, [email protected], Richard T Gretz, Stan Renard healthcare utilizations but the results are inconclusive. Moreover, many of the
reported results need to be interpreted with caution because of the self-selection
Terror Management Theory literature has focused almost exclusively on one issue. In this study, we propose an empirical approach which jointly models the
dimension of mortality - ‘uncertainty’ surrounding death. However, we postulate decision to obtain a general health check and healthcare utilization, tackling the
that there are additional dimensions, including ‘time scarcity’. We predict that self-selection problem by using the eligibility to obtain the health check for free as
when the ‘uncertainty’ dimension is salient, customers are more willing to make an instrumental variable. The eligibility has some exogenous variations by design
a material purchase compared with an experiential purchase, and vice versa and this helps us to partial out the effect of general health checks from the self-
when ‘time scarcity’ is salient. We argue that diverging effects are mediated by selection biases. We apply the proposed model to a large 12-year panel dataset
need for stronger material relationship and perception of ‘scarce-self’ provided by the Korean National Health Insurance Service. We find that
correspondingly. We test our hypotheses using a unique data set on the music participation in the general health check increases healthcare utilizations and
industry covering daily album sales from January 1, 2015 through December 31, ignored self-selection generates substantial upward bias in the estimates. We also
2017 for all artists in the billboard top 200 and all artists who passed away during find that the effect largely varies with several key demographic variables such as
100
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 101
n SC13
n SC12 Room 746, Alter Hall
Room 745, Alter Hall New Products and Innovation IV
Mobile Promotions Contributed Session
Contributed Session Chair: Ho Kim, University of Missouri-St. Louis, College of Business
Administration, 1 University, St Louis, MO, 63121, United States,
Chair: Ping Zhao, Wilfrid Laurier University, P2078 Peter’s Building, 75 [email protected]
University Ave W, Waterloo, ON, N2L 3C5, Canada, [email protected] 1 - Giving Green to Get Green? Yes and No
1 - Maximizing Customer Lifetime Value through Strategic Channel Cheng He, Georgia Institutes of Technology, 800 W Peachtree
Management: How to Incentivize Customers to use Apps Versus Street, Atlanta, GA, 30380, United States, [email protected],
Websites O. Cem Ozturk, Naiqing Gu
Gokhan Gecer, Doctoral Candidate, University of Mannheim,
The creation of a sustainable community through the adoption of green
Mannheim, Germany, [email protected],
technologies has become an important topic for the society in the presence of
Florian Kraus, Florian Stahl global warming. Federal and local governments have tried to promote green
The mobile channel is different from other channels since at least one of the technologies through various incentive programs (e.g., HOV access privilege or
engaged parties isn’t at a fixed location and these parties can communicate at any tax credit). Research in the past decade has provided divergent views on the
time (Balasubramanian et al. 2002). The unique nature of mobile devices leads effectiveness of such programs. In this paper, we empirically study the impact of
customers to develop habitual interactions with firms (Wang et al. 2015) and as a government incentives on the adoption of green vehicles in the U.S. automobile
result, the customer lifetime value (CLV) of mobile channel users tends to become industry. To identify this effect, we take advantage of incentive changes in certain
larger than the CLV of other channel users. One of the main reasons is that users’ states and we employ a difference-in-differences approach. Our findings suggest a
order rates increase when they adapt to mobile apps, resulting in a higher net significant positive impact of incentives on the sales of green vehicles. We also
monetary value in total spending (Naarang and Shankar 2016, Wang et al. 2015). find heterogeneous consumer responses to different incentive programs.
Consequently, online retailers should encourage the customers to switch to Furthermore, we identify the impact of incentive programs on non-hybrid
mobile devices by offering a discount over mobile apps. The purpose of this article vehicles to investigate whether the change in sales for hybrid vehicles is due to
is to develop a modeling approach to answer the research questions below and market expansion or substitution from non-hybrid vehicles. Our findings shed
test the proposed model in a field experiment conducted with a commercial light on the effective use of government resources to encourage green technology
website. 1. To what type of customer should a discount over the mobile app be adoption.
offered?2. What is the optimal discount rate? Firstly, we conducted a lab
2 - Computer-aided Exploration Of Product Designs in
experiment to test the assumptions used for constructing the analytical model. In
the model, we apply to discounted expected transactions (DET) (Fader et al. High-dimensional Visual Spaces
2004) to calculate the CLV of a customer. It is based on the customer purchasing Yegor Tkachenko, Columbia Business School, 3022 Broadway,
probabilities over different sales channels. We compare the expected transactions 5R Uris Hall, New York, NY, 10027, United States,
in case a discount over the mobile app is offered versus the transactions if the [email protected], Asim Ansari, Olivier Toubia
discount is not offered. Our preliminary experimental and analytical results
Recent computer science research has demonstrated that deep generative models
provide us with insights on 1) how the components of the model change the
can be used to extract lower-dimensional representations from rich image data
DETs, 2) where the thresholds for “discount offers” can be expected, and 3) what
and to regenerate high-quality images from such lower-dimensional
the optimal discount rate should be. Based on our results, we derive various
representations. We argue that these deep learning techniques can be effectively
implications for researchers and managers.
used for computer-aided exploration of visual product designs, where alternative
2 - Targeted Coupons on a Heterogeneous Customer Base design methods, such as conjoint or brute-force search, may not be applicable or
Arun Gopalakrishnan, Assistant Professor, Washington University may perform suboptimally. In this work, we first confirm properties of the lower-
in St Louis, Olin Business School, One Brookings Drive, St Louis, dimensional (latent) space that are desirable if we are to use it to generate novel
product designs. Specifically, we show (a) how distance between images in this
MO, 63130, United States, [email protected], Young-Hoon Park
latent space mimics human similarity judgments about the actual images, and (b)
Coupons targeting customers based on their value are common in practice as an that important characteristics of interest, such as product prices, can be predicted
efficient means of influencing consumer search and purchase behavior. Little from latent image data alone. Building on these findings, we propose and
research has examined why these coupons work and how the effects vary for a demonstrate machine learning techniques for exploration and ideation in the
heterogeneous customer base. We evaluate the causal impact of targeted coupons design space, such as image interpolation to generate product designs that are
on consumer search, redemption, and purchase behaviors using data from a similar to competitor’s products, or constrained Bayesian optimization to find
large-scale field experiment in which low- and high-value customers are exposed novel designs that score high on quantitative characteristics of interest. We use
to targeted coupons of different monetary values. We find that (1) coupons images and attributes of products sold on Amazon.com as well as human feedback
increase average revenues due to the increase in purchase incidence that does not from Amazon Turk workers as a basis for our experiments. The results of our
involve coupon redemptions; (2) coupons with higher monetary value increase work suggest that deep generative models offer a promising avenue for partial
101
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 102
102
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 103
104
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 105
105
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 106
n SD09 n SD10
Room 239, Alter Hall Room 605, Alter Hall
Reviews & Social Environments Digital Strategy
Contributed Session Contributed Session
Chair: Anita Luo, Georgia State University, Department of Marketing, Chair: Tolga M Akcura, Ozyegin University, Kusbakisi Caddesi No 2,
35 Broad Street NW, Atlanta, GA, 30303, United States, [email protected] Altunizade Üskÿdar, Istanbul, 34662, Turkey, [email protected]
1 - Residential Mobility Privacy and Risk Seeking Behavior 1 - Identity Fragmentation Bias
Ruijuan Wang, Doctoral Student, Wuhan University, 299 Bayi Tesary Lin, University of Chicago Booth School, Hyde Park, 5807
Road, Wuhan, 430072, China, [email protected], Jian Ni, Woodlawn Avenue, Chicago, IL, 60637, United States,
Fue Zeng [email protected], Sanjog Misra
This paper examines the effect of residential mobility on risk-taking behaviors in It is increasingly common for companies to collect customer data via different
different societal situations. Using both field and experimental data, we show that channels and sources. However, such data often contains different identifiers for
the effect of residential mobility on insurance purchase is moderated by privacy the same individual. Failure to consistently identify the same customer leads to a
concern, depending on the social exchange (social sharing) intentions among fragmented measure of exposures and behaviors, causing biased estimates and
decision makers. Consumers who live in a high-mobility community, are mainly erroneous inferences. Contrary to conventional wisdom, we show that the
motivated by self-enhancement (e.g., self-expression achievement, aloneness), estimation bias does not always take the form of attenuation; instead, the
and this self-oriented incentive fosters risk-seeking behaviors. While this effect estimator can have an arbitrary magnitude compared to the true value, making
disappears if residents are sensitive to privacy concern being unwilling to share interpretation of estimates difficult. Bias persists even when the goal of analysis is
sensitive personal information. Conversely, consumers who live in a stability obtaining a simple average effect. Using an analytical framework, we show that
community, are mainly driven by emotional-connection (e.g., social acceptance the estimation bias consists of two parts: a level shift caused by within-person
and dependence), and this stableness-oriented mindset leads to risk-aversion. The activity bias, and a magnitude change caused by missing links, which can be
provincial-level insurance data across China suggest that residential mobility further decomposed into measurement error and omitted variable bias.
foster financial risk-seeking behavior. We then analyze a rich individual-level Attenuation bias occurs only under certain symmetry condition, which we
insurance purchase dataset and test the potential mechanism of privacy concerns. formally characterize. We propose two different solutions to the estimation
Finally, in the control experiments, we manipulate residential mobility and problem caused by identity fragmentation, which are tailored to different stages of
privacy concern, and demonstrate that the “mobility” type participants’ risk- the data collection process. First, careful experiment design and analysis permits
taking behavior depends on their states of privacy concern and are is sensitive in estimator de-biasing, provided that the focus is on intent-to-treat. Second,
risk-seeking related tasks (i.e., investment decisions), whereas the “stability” type stratified aggregation is generally useful when data is already generated, and is
participants are more sensitive in risk-aversion related task (i.e., insurance suitable for a wider range of analysis. We discuss potential complications for
decisions). inference when data is linked by a third party company. Our findings have
important implications for experiment design, cross-channel analysis, and data
2 - Examining the Cost of Over-solicitation in Direct Marketing linking practices.
Anita Luo, Georgia State University, Department of Marketing, 35
Broad Street NW, Atlanta, GA, 30303, United States, [email protected] 2 - The Strategic Role of Featured Product Recommendations on
Online Retailers Competition
Consumer privacy related to direct marketing is an important topic that has
Skander Esseghaier, Associate Professor, ESADE Business School,
received attention from the public and business alike. Goodwin (1991) defines
consumer privacy based on control of information and control over unwanted Avda. Torre Blanca 59, Sant Cugat, Barcelona, 08172, Spain,
intrusion into the consumer’s environment, which includes unsolicited direct mail [email protected], Sofien Bahani
and telemarketing. Shiman (1996) suggested that the sheer volume of email Online retailers such as amazon.com and barnesandnoble.com make extensive
solicitations could cause consumer concerns as it could be over-used for its low use of product recommendations in categories such as books or movies, where
cost. Therefore, this research explores the issue of over-solicitation in direct they essentially carry the same assortment of products. Consumers in these
marketing and whether companies might actually incur financial loss as a results markets differ in their store preference (“loyal” vs. “switchers”) as well as in the
of perceived privacy intrusion. Although there is extensive literature that has strength of their product preference (“sure” vs. “unsure”). It is therefore puzzling
examined consumer privacy, no empirical studies have been conducted to that online retailers still rely on a traditional one-size-fits-all approach to product
examine the cost of over-solicitation in direct marketing. To address this issue, this recommendations, recommending the same “featured” product (e.g., “editor
study models the amount of direct mail sent, the number of emails sent, the spotlight pick” or “best seller of the month”) to all their customers in the category.
number of emails opened given emails sent, and the purchase amount jointly We examine the strategic implications of recommending a single featured product
through vine copula. The model is applied to a U.S. retailer selling multiple in the presence of such consumer heterogeneity. We show that such a
categories of home furnishing goods, which frequently sends out direct marketing recommendation approach enables differentiated retailers to mitigate the
communications to its customers. By examining how a high frequency of email intensity of price competition. Specifically, these recommendations steer store
and direct mail sent impacts customers’ purchase behavior, we are able to loyal consumers toward the featured product, increasing the size of store loyal
examine the cost of over-solicitation in direct marketing. This study provides purchasers (relative to switchers) in the featured product market. This, in turn,
empirical guidance to help companies examine their direct marketing programs so reduces a store’s incentive to compete aggressively for switchers in that market. In
they don’t waste valuable marketing resources and actually alienate their equilibrium, both stores feature the same product, compete less aggressively in
customers from over-solicitation. the featured product market, and generate higher profits. In an online context,
3 - How Dispersive Opinions Affect Evaluation: Edowment Effect however, stores have substantial information about their loyal customers and are
able to develop a preference-based approach to recommendations. Different
Guides Attribution Choice
consumers would be recommended different products that best match a store’s
Guangming Xie, Doctoral Candidate, Southwest Jiaotong information about their customers’ individual preference. Interestingly, we find
University, No. 111, North 1st section, Second Ring Road, that in a competitive context, such an approach may not benefit the stores. We
Chengdu, 610031, China, [email protected], show that even if it were costless for stores to develop such preference-based
Pei Hu, Yushi Jiang recommendation systems, they would be better off with a one-size-fits-all
approach, recommending the same featured product to all their loyal customers.
We show that the attribution choice of word-of-mouth (WOM) dispersion is
This result challenges the widely held belief that the more information a firm has
systematically influenced by the magnitude of endowment effect for the product
on its customers, the better off it would be.
or service: high endowment-effect results in greater weighting of reviewer-
related attribution (i.e., tastes, personalities, and individual styles) and reduced
weighting of product-related attribution (i.e., look, quality, and texture). Three
studies extend existing attribution-based approach to underscore the role of a
contextual variable (i.e., endowment effect). Individuals will be more likely to
choose a high-dispersion option over a low-dispersion one for a product or service
for which they exhibited a strong endowment effect. As an important implication
of our work, we also highlight a critical point that the mediating role of
attribution choice in the relationship between WOM dispersion and purchase
intention is moderated by the endowment effect. These findings reveal an
important boundary condition of traditional models of consumer attributional
inference and respond to the longstanding call for further exploration of the role
that social distributions play in consumer decision-making process.
106
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 107
107
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 108
108
Marketing Science back matter_Marketing Science 2018 5/29/18 2:53 PM Page 109
109