Received: 24 April 2020 Accepted: 20 July 2020

DOI: 10.1002/joom.1115

JOM FORUM

The lenses of lean: Visioning the science and practice

of efficiency

Wallace J. Hopp1 | Mark S. Spearman2,3

1
Ross School of Business, University of
Michigan, Ann Arbor, Michigan
Abstract
2
Strategic Project Solutions, Inc., San There is a significant gap between the descriptions of Lean used by industry
Francisco, California practitioners and the various bodies of academic research that have studied
3
Project Production Institute, San the theory and application of Lean. There is also a gap between applied
Francisco, California
research on Lean and basic research in the mathematical, physical and social
Correspondence sciences. As a result, Lean practice is based largely on trial-and-error experi-
Wallace J. Hopp, Ross School of Business, ence while potentially valuable research results remain locked away unused in
University of Michigan, Ann Arbor,
MI 48109. archival journals. This paper attempts to close these gaps by describing four
Email: [email protected] “Lenses of Lean,” each of which aligns with a practical perspective and rests
on a distinct body of conceptual research. Our hope is that this framework will
Handling Editor: John Gray
provide a useful construct for Lean training and implementation and will also
spur academic research that is relevant to advancing Lean practice.

KEYWORDS
efficiency, factory physics, lean, operations management

1 | W HA T I S L E AN ? Just as Lean was rooted in TPS, TPS was rooted in an

earlier system introduced at Ford (Ford, 1922). The Ford
The term “Lean” was coined by Krafcik (1988) and popular- Production System revolutionized manufacturing by
ized by Womack, Jones, and Roos (1991). But despite its introducing the moving production line to complex prod-
undeniable success as a label, “Lean” is actually a pretty old ucts like automobiles. It also established a focus on effi-
name for a very old concept. To begin with, Lean was cient, waste-free flow that remains at the center of TPS
explicitly derived from the Toyota Production System (TPS). and Lean systems. The Henry Ford Health System still
Indeed, the original descriptions of Lean equated it with uses the title Henry Ford Production System for their col-
TPS. The seminal book on Lean, The Machine that Changed lection of practices that include Kanban, just-in-time,
the World, stated “Toyota had fully worked out the princi- standardized work and other tools right out of TPS
ples of lean production by the early 1960s” (Womack (Zarbo & D'Angelo, 2006).
et al., 1991, p. 68). The first paper on TPS in English Of course, Ford did not create his system from whole
appeared in the late 1970s (Sugimori, Kusunoki, Cho, & cloth either. Moving assembly lines existed before the
Uchikawa, 1977), but it took until the 1980s for real interest Highland Park plant began using one to produce Model
to be sparked by descriptions of TPS in popular books T cars in 1913. Efficiency methods in management can
(Hall, 1983; Schonberger, 1982). By the 1990's, these had be traced backward through the Scientific Management
been digested and the Lean label had begun promoting the movement around the turn of the 20th century, the
application of TPS beyond the automotive industry. These American System of Manufacture in the 19th century,
led to extensions in the definition and tools of Lean, to the the (First) Industrial Revolution in the 18th century, the
point where now TPS is considered Lean but not all Lean European Scientific Revolution in the 16th and 17th cen-
implementations are considered TPS. turies, engineering and manufacturing innovations in

J Oper Manag. 2020;1–17. wileyonlinelibrary.com/journal/joom © 2020 Association for Supply Chain Management, Inc. 1
2 HOPP AND SPEARMAN

China during the first Millennium, and back at least as evolution and the entire discipline of Operations
far as the construction of the Pyramids of Egypt begin- Management, is a simplification of reality, and
ning before 2,500 BC. For as long as humans have been hence “wrong,” we can accept the pragmatic view
producing goods and services, making efficient use of that any definition of Lean that elevates practice or
resources has been a priority. deepens our understanding of how to elevate prac-
The fact that Lean is not new does not diminish its tice is good.
importance. Enhancing efficiency increases productivity,
which in turn elevates standards of living. Hence, it is With these in mind, we first describe what we feel is
not an overstatement to say that efficiency, the goal of the broadest defensible definition of Lean, which delin-
Lean, is central to improving the human condition. eates the potential scope of Lean research and practice.
But what exactly is it? The Lean Enterprise Institute We then offer four narrower definitions of Lean that
defines the core idea of Lean as “maximize customer reflect the range of descriptions from the research and
value while minimizing waste.” This definition is appeal- practitioner literature. Each, while admittedly incom-
ingly simple but decidedly unhelpful for implementation plete, points toward a connection between practice and a
or research purposes. The list of tools associated with particular body of research. Each is also appropriate as a
Lean and TPS (e.g., Kanban, 5S, kaizen, andon, heijunka, practice guide in particular environments. This allows us
jidoka, etc.) contains many useful ideas but also falls to leverage these four definitions to describe four lenses
short of a unifying definition. through which Lean can be viewed productively and
To produce a working definition (actually, multiple which provide the framework for this paper.
definitions) of Lean that will serve our goal of bridging The maximal definition of Lean follows from examin-
the gap between research and practice, we begin with ing what part of the Operations Management
three observations: (OM) discipline it comprises. Obviously, if Lean encom-
passes all of OM it is redundant and unnecessary. Con-
1. History did not start with Toyota. Lean may have versely, if Lean refers to a narrow set of practices in a
started out as a generic label for TPS, but it but its limited set of environments, it is not very interesting from
roots predate TPS and its applications go well beyond a scholarly standpoint or important from a practice
it. Limiting Lean to the philosophy or practices at standpoint. Therefore, we aim for something in the
Toyota is arbitrarily narrow and unsuited to our middle.
purposes. We do this by noting that a mainstream definition
2. There is a difference between a title and a term. A title of Operations Management is the business function
is simply a name that can be applied however one responsible for effective delivery of goods and services
likes, but a term requires a definition. In the world of to customers. If we define goods/services and cus-
practice, Lean is generally used as title, which can tomers broadly enough, virtually any organization is
refer to a collection of practices, a near-religious com- engaged in delivering goods and services to cus-
mitment to waste reduction, or just good operations tomers. As such, OM is fundamentally about facilitat-
management, rather than a well-defined term. ing execution of a business strategy. However, what
Although such usage is maddeningly imprecise from a this means in a given setting depends on how “effec-
scholar's perspective, titles can be useful in practice. tive delivery” is defined. Fundamental dimensions,
Henry Ford Production System, Michigan Quality Sys- along which customer concerns can be described,
tem and Danaher Business System are examples of include quality, time, variety and price/cost. Manage-
titles for collections of ideas and practices without rig- ment of each of these defines an OM sub-field. For
orous boundaries or definitions, but which serve to example, the quality dimension is the focus of the
focus and motivate activities within an organization. quality management area of OM, which has been
However, because Lean is used as a loosely specified addressed in practice under the headings of Total
title in practice, scholars that attempt to use it as a Quality Management and Six Sigma. The time dimen-
term are usually vague (when they fail to define Lean) sion is the focus of the responsiveness management
or quarrelsome (when they do and then take on all theme of OM, which has been addressed in practice as
the writings that conflict with their definition). Time-Based Competition, Agile Manufacturing, and
3. All models are wrong, but some are useful. This apho- Quick Response Manufacturing (Suri, 1998). Variety
rism, usually attributed to statistician George Box, is a key focus of the product management area of OM,
offers a way out of the vague vs. quarrelsome trap. If which has been addressed in practice as Product Port-
we admit the reality that every conceptual framework, folio Management and (in part) by Material Require-
including statistical mechanics, the theory of ments Planning.
HOPP AND SPEARMAN 3

Almost everyone would agree that Lean focuses pri- not comprehensive, but it spans the gamut of ways, from
marily on efficiency, which means it addresses the price/ simple to sophisticated, that Lean has been described by
cost dimension. Although one could further limit Lean scholars and practitioners:
within the efficiency space to certain environments
(e.g., systems involving repetitive production or limited 1. Lean is the pursuit of waste elimination: This is the
amounts of variability), this seems arbitrary to goal implied by the Lean Enterprise Institute defini-
us. Therefore, for the purposes of this paper, we will tion and is the essence of most descriptions of Lean in
equate Lean with efficiency management and regard any- the practitioner literature. Some of these emphasize
thing that increases the efficiency of delivering products value in place of waste, for example by stating the goal
as a Lean practice. Under this definition, Lean is analo- as “creating more value for customers with fewer
gous to but distinct from quality, responsiveness and resources” (Lean Enterprise Institute, 2019). But, since
product management in the OM spectrum.1 Lean prac- waste is defined as anything that does not create value
tices, such as value stream mapping or Kanban are analo- for customers, it is essentially the same as reducing
gous to specific practices within these areas, such as waste. While indisputable as a goal, this definition
fishbone diagramming and statistical process control in only barely goes beyond the overarching efficiency
the quality management area. management definition by introducing the concept of
We acknowledge that this broad definition begs the waste. Because it offers no explicit guidance on the
question of whether we need Lean as a term at all. Why causes or cures of waste, this definition tends to foster
not simply call it efficiency management? Why not a focus on the most directly visible forms of waste and
indeed. Using industry titles, which are imprecisely straightforward ways to eliminate them.
defined and subject to change, as research terms is 2. Lean seeks to minimize the cost of excess inventory,
fraught with problems. For instance, consider the case of capacity or time: This definition invokes a basic
Six Sigma, which, like Lean, is a title that was applied to insight from Factory Physics that variability must be
emergent industry practices. Six Sigma started out as a buffered by some mixture of inventory, capacity and
defect measurement tool at Motorola, where it was grad- time (Hopp & Spearman, 2008). The implication is
ually expanded to include other statistical quality control that variability is the main cause of waste beyond the
tools. At that stage, it was clearly a methodological subset obvious direct waste inherent in unnecessary activi-
of the quality management area. But when GE adopted ties. By emphasizing the cost of variability buffering,
Six Sigma and added DMAIC, training with “belt” certifi- this definition forces a focus on both variability reduc-
cation levels, and other features, it grew into a full- tion and finding the most efficient mix of buffers for
fledged management system that could be equated with the variability that remains (Hopp &
quality management as a whole. Even if one does not Spearman, 2004).2 Framing Lean in terms of variabil-
regard the GE version of Six Sigma as encompassing all ity buffers highlights the underlying causes of waste
of quality management, it is only a matter of time until and thereby helps to discover less obvious sources of
another firm defines it as such. Indeed, it is now common waste than those likely to be surfaced by the first
to find descriptions of Six Sigma that include TPS prac- definition.3
tices, which takes it even beyond the quality space into 3. Lean is a systematic process for reducing the cost of
the efficiency space. Such shape shifting definitions make waste: Unlike the first two definitions, which define
communication difficult and rigorous research impossi- Lean in terms of its goal, this definition focuses explic-
ble. Consequently, we would prefer to use generic terms itly on the improvement process. There are many
like quality management and efficiency management, books and articles in the practitioner literature that
which can be rigorously defined and will always be with take this approach, generally by starting with the
us, rather than imprecise industry labels like Six Sigma basic “production without waste” definition and then
and Lean that will eventually be replaced by trendier laying out a series of steps for operationalizing it. For
titles. example, Womack and Jones (1996) described a five-
But, for the purposes of this paper, it does not really stage sequence consisting of Define Value, Map the
matter whether we use Lean or efficiency management. Value Stream, Create Flow, Establish Pull, Pursue
Either way, we are still only providing a heading for the Perfection. By focusing on the improvement process,
problem area without any hint at structure or solution this approach to defining Lean promotes a more
approaches. This is why writers in the academic and expansive system-wide perspective and a management
practice literatures have offered definitions of Lean that focus that is absent from the first two definitions.
go beyond simply categorizing it as focusing on effi- 4. Lean is an organizational culture that encourages con-
ciency. The following list of Lean definitions is certainly tinual reduction of the cost of waste: This perspective
4 HOPP AND SPEARMAN

takes us fully into the realm of management by recog- studies of manual work (Gilbreth, 1911). In addition to
nizing that it is people working within organizations establishing the concept of “standardized work” that
that create and manage the systems for carrying out would become central to TPS and Lean, the Scientific
the waste elimination activities called for in Lean. Management movement lead to creation of the field of
Spear and Bowen (1999) and Liker (2004) are exam- Industrial Engineering (IE).
ples of practice-oriented descriptions of Lean that However, because the “pursuit of waste reduction”
focus on creating cultures of waste elimination. For definition of Lean lacks any reference to underlying cau-
the most part, however, such descriptions are attempts ses of waste, it basically leaves the user to focus on waste
to summarize characteristics of Toyota and other suc- that is directly visible from observation of individual pro-
cessful Lean organizations, rather than applications of cesses. For this reason, we term the lens implied by this
behavioral and organizational science to guide the definition the Process Lens. Taylor and Gilbreth focused
creation of a successful Lean culture. on waste at the process level in their work during the Sci-
entific Management era in the early 20th century. For
We reiterate that none of these definitions is “correct” example, the waste on which Taylor focused his time
in any rigorous sense. But, in the spirit of “all models are studies (e.g., shoveling with an improperly sized shovel)
wrong, but some are useful,” each has practical value. By is appropriately termed processing waste. In contrast, the
providing four different “lenses” through which Lean can waste at which Gilbreth aimed his motion studies is logi-
be viewed, these suggest modeling frameworks for cally termed motion waste, since it involves unnecessary
scholars and problem-solving perspectives for practi- motions (e.g., excessive reaching and lifting bricks). Con-
tioners. In the sections that follow, we examine each of temporary descriptions of Lean have added five more cat-
these lenses, trace them to the relevant bodies of egories of waste to those addressed by Taylor and
research, and describe the environments in which they Gilbreth to produce following list of seven types of waste:
are likely to be most useful in identifying the causes and
remedies for waste in production and service systems. We 1. Defects
will argue that the simple lens based on the first defini- 2. Overproduction
tion has been used (overused?) in practice but provides a 3. Transportation
limited basis for research, that the lenses based on the 4. Waiting
second and third definitions have been the basis for the 5. Inventory
most important research into Lean but may not have 6. Motion
been as fully exploited by practitioners as they deserve to 7. Processing
be, and that the lens based on the fourth definition is ripe
with opportunity for scholars and practitioners alike. Although this set of waste categories is ubiquitous in
Furthermore, when taken together, this set of Lean the Lean literature it is neither precise nor valuable. For
lenses highlights opportunities for linking disparate example, an integrated steel mill that can only produce
streams of research to yield deeper insights into the sci- steel in 250-ton batches (or “heats”) will have leftover
ence of efficiency and also suggests a staged structure for slabs any time a customer orders less than 250 tons.
Lean implementations. These will pile up in storage waiting for another order for
that particular blend of steel. Is this overproduction
waste? Inventory waste? Does it matter? We argue that
2 | THE PROCESS LENS the label is irrelevant to the elimination of the waste.
What matters is that the large and rigid batch size makes
The first definition of Lean is “pursuit of waste it impossible to match supply with demand efficiently.
elimination.” This goal has been behind every effort to Reducing this waste—whatever it is called—requires
develop a better way to produce a good or service since changing the process to facilitate smaller and more flexi-
the dawn of civilization. But it was not formalized until ble batch sizes or finding less costly ways to accommo-
the late 19th and early 20th century under the banner of date the current batch size. Unfortunately, nothing in the
Scientific Management. Frederic W. Taylor, the Father of Process Lens or the list of waste categories it has engen-
Scientific Management, advocated “one best way” to per- dered guides us to this conclusion.
form a given work function and introduced the practice Because it lacks guidance on how to diagnose and
of time studies to discover such best practices systemati- remedy waste, the Process Lens is best suited to identify-
cally (Taylor, 1911). Fellow Scientific Management pro- ing waste that is created directly in the process itself,
ponent Frank Gilbreth carried the process for finding rather than as a by-product of issues or activities outside
best practices a step further through detailed motion the process. As such, it can help us identify ways to
HOPP AND SPEARMAN 5

improve efficiency by improving specific steps in a pro- Fluctuations in customer preferences, order sizes, sea-
cess. Such improvements can be facilitated by a classic sonal trends and many other factors contribute to
industrial engineer (with a clipboard and stopwatch) or demand variability, while fluctuations in production rate,
by workers with sufficient training to do classical IE anal- product yield, rework rate, staffing, delivery times and
ysis of their operations. For example, a 5S reorganization many other factors contribute to supply variability.4 Vari-
of a workstation on an assembly line can be very effective ability may or may not be accompanied by uncertainty.
in revealing and reducing motion and transportation For example, planned shifts in production output consti-
waste. For organizations just beginning their Lean jour- tute variability but are not uncertain, while unplanned
ney, this very simple view of waste reduction can be help- reductions in productive capacity due to equipment fail-
ful in attacking obvious waste. But this lens is less ures result in variability that is also uncertain.
helpful in identifying waste that propagates from other Omitting variability from discussions of Lean, as is
parts of the process or beyond it. It is also not terribly often the case in descriptions based on the Process Lens
useful in enabling organizations to drive out waste on a perspective, is a huge oversight because (a) all real-world
continual basis because it does not provide any frame- systems involve variability and (b) addressing indirect
work for prioritizing different sources of waste. Firms waste caused by variability is more difficult than
seeking to go beyond the most basic level of Lean need a addressing direct waste from unnecessary activity. Man-
more detailed framework than the Process lens. aging production without variability is like farming with-
The Process Lens is also limited as a frame for out weather: It would be easy if it existed. Just as
research. Since it lacks any perspective on the means for managing the impacts of weather fluctuations is essential
eliminating waste, it cannot draw on bodies of basic to farming success, managing the indirect waste due to
research for guidance. This leaves direct research into variability is fundamental to business success.
Lean as the only avenue for research suggested by the The fundamental law of Factory Physics that
Process lens. For example, scholars have looked for describes how variability causes waste can be stated as
empirical evidence of the impact of Lean practices on follows (Hopp & Spearman, 2008, p. 309):
firm performance (Belekoukiasa, Garza-Reyesb, &
Kumarc, 2014; Lopes Negr~ao, Godinho Filho, & Variability-Buffering Law: Variability in a
Marodin, 2017; Moraro, Lemstra, & Nwankwo, 2016) and production system will be buffered by some
on the workforce (Parker, 2003; Ulhassan, von Thiele combination of inventory, capacity and time.
Schwarz, Thor, & Westerlund, 2014; Vidal, 2007). But to
go further and learn from research other than direct stud- To make the concept of buffers concrete and to show
ies of Lean systems, we need a richer definition of Lean. how they relate to the various types of waste, we consider
a sandwich station within a retail bakery. Demand for
sandwiches is variable due to factors such as time of day,
3 | T H E FL O W LE N S weather, and the whims of individual customers. Supply
(i.e., the pace at which customers are served) varies due
The second definition of Lean is “to minimize the cost of to product differences (e.g., a hot sandwich takes longer
excess inventory, capacity or time.” The explicit identifica- to make than a cold one), stockouts (e.g., the sandwich
tion of exactly three types of buffers is essential to Lean station runs out of Provolone cheese and the clerk needs
because the core challenge in delivering goods or services to go in back to slice more), and other factors. As a result,
to customers is efficiently matching supply with demand. there will be intervals in which the demand rate exceeds
The Process Lens focuses on the efficiency element of this the supply rate and a backlog of customers builds up.
challenge by seeking to drive out waste in execution cau- The queue that results when supply lags demand con-
sed by activities that do not add value to the customer. stitutes a time buffer since it involves customers waiting
While this is fine for highlighting waste due to unneces- for their sandwich. The most straightforward way to
sary motion or process errors, it cannot help us identify reduce this customer waiting is to put more clerks on
waste that is the result of variability in the supply and sandwich making duty. The excess capacity of the clerks
demand processes. The second definition of Lean zeroes during slow periods represents a capacity buffer since it is
in on variability-induced waste. Because the effects of capacity held in reserve to deal with demand spikes. An
variability are manifested in the flows of people, mate- alternative for reducing customer waiting is to make up
rials, dollars or other entities, we term this second lens common sandwiches (e.g., turkey and swiss on white) in
the Flow Lens. advance. These stocks of pre-made sandwiches, which
Variability is defined as any deviation from absolute will not stay fresh as long as the individual ingredients
regularity (see Hopp & Spearman, 2008, Chapter 8). and hence are at risk of being discarded, represent an
6 HOPP AND SPEARMAN

inventory buffer. The more variability in demand and/or diminishing returns when adding buffers. This, in turn,
supply, the more inventory and/or capacity and/or time implies an optimal buffer configuration for almost any
buffering there must be. cost structure for the buffers.
In the language of the seven types of waste from the It is important to note that, as emphasized by under-
Process Lens, time buffering is a form of waiting waste lining in the Variability-Buffering Law, these buffers are
and inventory buffering is a form of (what else?) inven- not optional. If there is variability, there will be buffers.
tory waste. The other five types of waste (defects, over- A small capacity buffer will lead to larger fluctuations of
production, transportation, motion, processing) could be net-inventory that create either larger stocks or more
capacity buffers if they are the result of variability backorders. The only way to reduce the total amount of
(e.g., an unexpected spike in demand leads to defects due buffering (as measured by the product of the buffers) is to
to rushing or extra motion to get an ingredient that has reduce total variability.
run out) or if they provide protection against future vari- It is worth noting that TPM practices include a host
ability (e.g., overproduction produces extra inventory that of variability-reduction techniques, such as, total quality
can be used to meet a demand spike). But they are often management, preventive maintenance, production
simple waste due to poor design or execution (e.g., a smoothing, standardized work, and others. Each of these
badly laid out station that forces clerks to walk around to techniques serves to smooth demand and supply flow so
do their work). But the point of the variability focus of that they can be matched more efficiently. Although
the Flow Lens is not to take us back to the waste labeling almost never described in this manner in the Lean litera-
approach of the Process Lens. Instead it is to help us trace ture, TPS can be thought of as a giant variability-
the causes of variability and provide a guide to minimiz- reduction machine.
ing the total cost of variability buffering. To provide more structure to the search for ways to
The Variability-Buffering Law indicates two candi- reduce the cost of variability buffers, it is helpful to char-
dates for reducing the cost of variability buffering: (a) acterize variability in terms of predictability and cus-
reduce variability, and (b) alter the mix of variability tomer value. Variability can be predictable (e.g., planned
buffers. To do this in a systematic way, we need to under- downtime for maintenance) or unpredictable
stand how variability buffers interact. (e.g., emergency outages). This distinction matters
We begin by noting that time and inventory buffers because predictable variability can be buffered only when
are clearly mirror images of the other. If a part is pro- needed, while unpredictable variability must be buffered
duced before its demand, it waits in stock as an inventory continuously. This implies that predictive analytics tech-
buffer. If it is produced after the demand, the customer niques that convert unpredictable variability into predict-
must wait, incurring a time buffer. Variability causes a able variability can reduce buffering costs. For example,
lack of synchronization between demand and production, processing data from wear monitors to reliably determine
and hence results in both types of buffers. But we can when a machine needs preventive maintenance and
adjust the balance between them via an inventory control thereby prevent random failures will reduce the amount
policy. By increasing the target stock level, we will of inventory buffering needed to maintain downstream
increase the inventory buffer but, because it lowers the flow during outages.
likelihood of stockouts, decrease the time buffer. The A second useful characterization of variability is in
mechanics of this time-inventory buffer have been well terms of customer value. Variability can be the result of
described in inventory control theory (see Zipkin, 2000 errors (e.g., machine failures), which have no customer
for an excellent summary). value, or the result of actions that produce customer
We can also describe the relationship between the value (e.g., providing product variety). The former are
time-inventory buffer and the capacity buffer. Specifi- simply targets for elimination, while the latter pose
cally, as shown in Spearman and Hopp (2019), the prod- tradeoffs to be addressed through management decisions.
uct of the time-inventory buffer and the capacity buffer is For example, an automaker can reduce variability and
a constant that is monotonically increasing in the vari- cost by limiting the number of vehicle variants. Ford
ability of both the demand and the supply processes. This famously did this by painting all Model Ts black. Modern
means, for example, that if we double the capacity buffer, car companies address this variability vs. variety dilemma
the time-inventory buffer will be halved. This inverse by building many different models on a single vehicle
relationship describes a “diminishing returns” to buffer- platform. Managing variability rooted in customer value
ing. Each time we double the capacity buffer, the can be complex.
absolute increase gets larger, but each time we halve the To translate this conceptual description of variability
time-inventory buffer the absolute decrease gets smaller and buffers into practice, we return to the bakery
and smaller. In economic terms, this relationship shows sandwich-station scenario. In this system, we could
HOPP AND SPEARMAN 7

reduce demand variability by having customers call in Synchronization of supply and demand variability
sandwich orders so they can be pre-made during slow makes it possible to reduce buffering because it is actu-
intervals. We could reduce supply variability by stocking ally only unsynchronized variability in these two pro-
the sandwich station more carefully to prevent disrup- cesses that creates the need for buffering. If capacity
tions due to stockouts. Reducing either demand or supply increased and decreased in exact proportion to demand,
variability will reduce the amount of buffering required customer orders would be filled just in time without
and hence the cost. excess inventory or capacity. Of course, this never occurs
Since supply variability is largely self-inflicted, the in the real world because we lack sufficient information
bakery can reasonably strive to reduce it to near zero to perfectly match capacity to demand on a real time
through better ordering and stocking execution. It could basis. However, we can use flexibility to partially syn-
also make demand variability predictable by requiring chronize capacity and demand. An example in the sand-
customers to pre-order their sandwiches or make it mini- wich station scenario would be to cross-train workers
mal by having them sign up for time slots (e.g., like a from other areas of the bakery, such as the bakers or ser-
dentist does). But both of these would be burdensome to vice counter staff to help make sandwiches when
the customer and hence competitive disadvantages. Con- demand spikes. By using the flexible capacity provided by
sequently, demand variability in this scenario, as in many these cross-trained people only when needed, we will
other service environments, has positive customer value. require less capacity buffering than if we could only
This means the bakery will always have to buffer assign sandwich specialists who would be on duty during
demand variability. The challenge is to find the most eco- idle periods as well as busy periods.
nomical mix of capacity (sandwich clerks), inventory Of course, cross-training is more than simply a
(pre-made sandwiches), and time (customer waiting). variability-pooling technique. It also impacts the worker
Although sufficiently detailed data to do a formal cost experience by changing the nature of jobs. This impact
optimization is probably unavailable, it is certainly feasi- can be positive or negative. Negative impacts of Lean
ble to implement a satisficing solution that sets a cus- practices have been documented in studies such as
tomer waiting-time target that will satisfy the market and Parker (2003) and Vidal (2007). But we observed a posi-
then find by trial-and-error a staffing/inventory mix that tive example in a consulting engagement in which we
achieves it. were seeking ways to increase the capacity of a very
We can make variability buffering more efficient by expensive bottleneck. As we often do, we put the ques-
invoking the following useful corollary to the Variability- tion to the workers themselves in a workshop. The
Buffering Law (Hopp & Spearman, 2008, p 313): workers in a non-bottleneck unit quickly came up with a
plan to swap positions with the bottleneck team after
Buffer-Flexibility Corollary: Flexibility their lunch break and before the bottleneck team took
reduces the amount of variability buffering their lunch break. This effectively shifted 30 minutes of
required in a production system. labor capacity per shift (for three shifts) from the non-
bottleneck to the bottleneck. And, much to our delight, it
There are two reasons for the above corollary: (a) flex- made the workers in both units happier to have a greater
ibility facilitates variability pooling, and (b) flexibility variety of work during the day. We will drill more deeply
facilitates synchronization of supply and demand into the connection between the physics and psychology
variability. of flow in the next section.
Variability pooling takes advantage of the statistical There is a time analogy to the capacity matching
property that a sum of random variables is less variable, made possible through cross-training. We came across an
percentage wise, than the individual random variables in example of this years ago in a manufacturer of custom-
the sum. An example of pooling in the bakery sandwich ized institutional cabinets. The manufacturer had a bro-
station scenario would be replacing three separate stocks chure that promised 10-week lead times for delivery of an
of turkey and swiss sandwiches, one with mustard, one order. However, because one of their competitors had
with mayo and one with both, with a single stock of tur- begun quoting 4-week lead times, the firm was looking to
key and swiss sandwiches bagged with a packet of mus- shorten their quotes. We suggested shifting to dynamic
tard and a packet of mayo so customers can add either or lead times that would take into account the work backlog
both. Because combined demand for the three types of at the time the order was received. Because 10 weeks was
sandwiches is more predictable than the demand for each chosen as a lead time, they were almost completely confi-
individual sandwich type, we can carry less safety stock dent they could meet, it meant every customer received a
in the pooled case than in the un-pooled case. This is an “worst case” quote, even when the work backload was
example of flexible inventory. light. By moving to dynamic quotes, the firm was able to
8 HOPP AND SPEARMAN

continue quoting times they were nearly certain to meet generally less expensive to install and maintain than
and quote an average lead time significantly shorter than physical buffers, and emergency response.
10 weeks.5 As in the cross-training case above that made Is usually even cheaper.6 If the objective is to mini-
use of demand information to adjust flexible capacity in mize expected cost of variability, the more costly mea-
relation to demand, the dynamic lead-time quoting policy sures should be used where the likelihood and
made use of demand information to adjust lead times in consequences of disruption are high.
relation to demand. As such, it was an example of a flexi- Figure 1 provides a schematic for mapping variability-
ble time buffer. response strategies to different scenarios. Short-term vari-
Finally, an aspect of variability that has been largely ability that is highly likely to occur in a given interval is
overlooked in both the practitioner and academics litera- suited to buffering with inventory, capacity, or time, pos-
tures on Lean is that variability occurs over different time sibly in combination with the flexibility strategies dis-
scales. Minute-by-minute variation in supply and demand cussed earlier. In the cinnamon chip example, this would
lead to queueing delays. Day-by-day or week-by-week translate into each of the franchised bakeries stocking
fluctuations in product mix can shift bottlenecks. Rare but enough chips at their stores to ensure a high service level.
extreme events can trigger crises. The Variability-Buffering Medium-term variability that is too rare to justify the
Law and its Buffer-Flexibility Corollary apply to vari- constant presence of a buffer but still likely enough to
ability no matter what the time frame. Hence, a well- occur to warrant advance attention can make use of a
designed Lean program should go beyond addressing contingency plan. For example, the bakeries could imple-
the minute-by-minute variability we usually think of ment an inter-store sharing agreement under which they
(when we think of variability at all) in the context of agree to lend cinnamon chips to one another in emer-
Lean. This requires thinking about the interactions gency situations such a store that has chips damaged by a
between the buffers used for variability at different roof leak.7 Long-term variability that is too rare to justify
time scales. even a contingency plan must rely on emergency
To see this, consider a set of franchised bakeries in a response. For the bakeries, this might involve seeking out
metropolitan area. These independently owned bakeries a backup cinnamon chip supplier who could provide
make use of cinnamon chips they each purchase from a chips if the primary supplier were disrupted. Of course,
common supplier. Most of the bakeries receive replenish- all of these responses are costly, which implies they
ment orders every other week. Because the holding cost should only be used where the benefits outweigh the
is relatively low and the cost of stocking out is high costs. If not, then no action (e.g., do not bake cinnamon
(because these chips are needed in some popular prod- bread and try to divert customers to substitute products
ucts), Newsvendor logic argues for ordering enough cin- like cranberry-walnut bread) may be the optimal choice.
namon chips to ensure they almost never run out. The variability-response mechanisms for different
But what does “almost never” mean? Surely it means time scales are very different, but they impact one
having enough to cover plausible demand spikes and another. For example, if a bakery holds more safety stock
the occasional wastage due to burned bread or other to cover routine variability, it will be more likely to be in
mistakes. But does it mean having an extra supply in a position to share if another bakery has an emergency. If
another room in case the regular supply gets destroyed the bakeries collectively hold more combined safety
by water from a roof leak? Does it mean having enough
stock to protect against a fire at the supply plant that
cuts off deliveries for 6 months? Probably not. Holding Buffering/Flexibility
High
so much safety stock of every important ingredient
would require a great deal of space and would hardly
qualify as Lean.
Likelihood Contingency Planning
The key tradeoff here is one of cost versus responsive-
Do
ness. Physical buffers such as inventory and capacity are Nothing
expensive because they must be maintained at a cost con-
stantly. But they are ready for instant deployment. Two
Low
strategies that are less expensive and less responsive are Emergency Response
contingency planning and emergency response. Contin-
Light Severe
gency planning involves anticipating events and defining Consequences
plans for dealing with them, while emergency response
involves building some institutional readiness for events F I G U R E 1 Addressing variability on different time scales.
without detailed plans. Contingency planning is Source: Adapted from Hopp (2011)
HOPP AND SPEARMAN 9

stock, they will all be able to keep making product for a behavioral perspective, which we introduce in the fourth
longer time while seeking a new supplier in the event Lean Lens.
their regular supplier is disrupted. The Flow Lens focus Semantics aside, the Flow Lens is of great value to
on minimizing the cost of buffering implies that we researchers because it describes Lean in a way that makes
should think about variability on different time scales, it amenable to the vast range of probabilistic modeling
and the interconnections between the different types of tools in the OM arsenal. It is also highly useful to practi-
buffers. tioners who have already taken elementary steps, such as
Note that this sort of layered response to variability using value stream mapping and 5S to root out obvious
on different time scales is not limited to production sys- waste and are now seeking to find and eliminate more
tems. An example of a service system is a hospital emer- subtle forms of indirect waste.
gency room. Normal demand variability is buffered by
staffing to accommodate a high percentile of the demand
distribution. Demand spikes above this level may be 4 | THE N ETWORK L ENS
addressed by calling in off duty staff members. Even
more extreme demand scenarios, due to a mass casualty The third definition of Lean is “a systematic process for
event for instance, may be dealt with by diverting some reducing the cost of waste.” Unlike the first two defini-
patients to other emergency rooms in the area. Again, a tions, which help us to enumerate types of waste, the
Lean strategy is one that seeks to minimizes the total cost focus on a systematic process forces us to think about
of buffering over time, which will require using and coor- where and how to reduce waste to achieve maximum cost
dinating different buffering mechanisms. efficiency.
If you have done much reading of the Lean literature, In a production line or a simple supply chain, the
the majority of which is based on the Process Lens, you number of choices may be sufficiently limited to allow
may be thinking that Figure 1 and the discussion of it are the Flow Lens to achieve much of the potential of Lean.
not Lean at all. Indeed, even serious scholars such as Yin, But most production and service systems consist of many
Stecke, Swink, and Kaku (2017) have been careful to dis- interconnected flows. Examples include a plant with
tinguish between Lean, which deals with fixed buffers in thousands of product routings that share processes and
low variability environments (e.g., inventory to maintain people, a supply chain in which producers use multiple
flow in assembly lines), and Agile, which deals with flexi- suppliers and suppliers serve multiple customers, and a
ble buffers in high variability environments (e.g., cross- service system (e.g., a hospital) that serves many types of
trained capacity to respond to demand spikes). But there customers (patients) that require different but over-
is no scientific reason to distinguish between Lean and lapping sets of resources. All of these can be represented
Agile on the basis of the amount of variability in the envi- conceptually as networks of flows, so we label this the
ronment and the type of buffers used. The Variability- Network Lens.
Buffering Law and the Buffer-Flexibility Corollary apply One approach for addressing complex flow networks
to both scenarios. And, as noted by Yin et al., a number is to separate and simplify the flows, and then apply the
of practices (e.g., cross-training and lead-time reduction) Process or Flow Lenses to identify improvement options.
are associated with both Lean and Agile. This is essentially what cellular layouts, supplier consoli-
However, as Yin et al. also note, one can certainly dis- dation and focused factories do. While appealing where
tinguish Lean from by focusing on strategic intent. If practical, it can sometimes be very expensive to break up
Agile is defined as the science of responsiveness, then it complex networks into separate cells or factories because
is indeed different from Lean, which we equate with the this can require duplicate capacity. Hence, to develop a
science of efficiency. But as long as we are focused on systematic process for reducing the cost of waste, we need
efficiency, then it should not matter whether variability a means for diagnosing and improving complex flow net-
is high or low or buffers are fixed or flexible. Browning works directly.
and Heath (2009) recognized the danger of an overly rigid The core concept for understanding the behavior
distinction between Lean and Agile, noting that Lean of networks of flows, and thereby identifying points
“may provide even greater value by incorporating some of maximum leverage, is that of a bottleneck, which
aspects of agile manufacturing.” Finally, however, even if we define as follows (Hopp & Spearman, 2008,
the cases represented in Figure 1 represent a continuum p. 231):
of Lean from a scientific standpoint, it could still be use-
ful to use different titles (Lean and Agile?) in industrial Bottleneck Definition: The bottleneck in a
settings as a way to focus attention on different aspects of production or service network is the resource
the system. Evaluating the utility of such titles requires a (node) with the highest long-term utilization.
10 HOPP AND SPEARMAN

Queueing theory tells us that bottlenecks will produce seasonal, subject to grow, disrupted by occasional shocks,
the majority of backups and delay (inventory and waiting or affected by other factors. Capacity may be altered by
waste). The intuition behind this is that the closer a learning, staffing changes, investments or other factors.
resource is to full utilization, the smaller the fluctuation As a result, bottlenecks may shift over time or even be in
in demand or capacity needed to overload the resource. constant flux. In such cases, static capacity calculations
Small and transitory overloads that happen with high fre- and queueing analyses may provide little useful
quency cause normal queueing behavior (e.g., stores of guidance.
work-in-process inventory in a factory or lines of patients This suggests that there may be a need for more
waiting in a hospital). Large and extreme overloads that dynamic representations of production and service net-
happen rarely cause major backups and shortages works that make use not only of bottleneck and queueing
(e.g., jet planes grounded in airports or people waiting for analysis, but also of real-time data and network-structure
emergency supplies in a hurricane). Regardless of time analysis to evaluate vulnerabilities and effectiveness of
scale, bottlenecks must be key focal points in a systematic improvement options. An example of research that
effort to reduce the cost of waste. makes use of a data-driven network approach to analyze
Furthermore, queueing theory also tells us the effect and improve a production or service system include
of utilization is nonlinear (exponential), while the effect Gokpinar, Hopp, and Iravani (2010), who used networks
of variability (as measured by squared coefficient of varia- to represent product designs and organizational commu-
tion) is linear. This implies that we should first look for nication patterns to identify waste inducing mismatches.
ways to reduce utilization (e.g., by adding capacity and/or An example of a real-world attempt to incorporate a
eliminating unnecessary demand, such as that from dynamic, system-wide perspective into flow management
rework) and then look for ways to reduce variability of a complex system is the electronic dashboard adopted
(e.g., by smoothing flow into bottlenecks and/or reducing by the Johns Hopkins Hospital (Martinez et al., 2018).
variability in bottleneck processes). Leveraging research from queueing, network theory, data
Finally, in addition to bottleneck and queueing logic, analytics and machine learning in dashboards like this
we need to assess costs and benefits of candidate may open opportunities for Lean execution that have so
improvements, as we would in development of any busi- far eluded systems as complex as those in global product
ness case. Taken together, these tools comprise the con- design systems and research hospitals. Because these
ceptual basis for the Network Lens of Lean. system-wide efficiency initiatives require much more
In the world of practice, this perspective has appeared sophistication to pursue than the direct waste elimination
in various ways. The best-selling book The Goal activities promoted by the Process Lens or the indirect
(Goldratt & Cox, 1986) highlighted the importance of bot- waste elimination via variability reduction and buffering
tlenecks but in the context of simple flows rather than in promoted by the Flow Lens, the Network Lens is gener-
complex networks. Schmenner and Swink (1998) brought ally the purview of advanced Lean practitioners.
the concepts of bottlencks and variability together in the
theory of swift, even flow, and Yin et al. (2017) applied
this theory to the Japanese seru system to describe how it 5 | THE O RG ANIZATION L ENS
can elevate both efficiency and responsiveness. Even
basic Lean tools, such as Value Stream Mapping and The fourth definition of Lean is “an organizational cul-
Fishbone/Ishakawa/Cause-and-Effect/Five Whys dia- ture that encourages continual reduction of the cost of
grams, which are useful in breaking down and assessing waste.” The focus on organizational culture forces us
complex environments, are more effective when paired beyond the physics of flows that were at the center of the
with the bottleneck/queueing/cost–benefit analysis per- first three Lean Lenses. Although a physics focus might
spective of the Network Lens. be sufficient for a production or service system run
One reason for the gap between the conceptual per- entirely by machines (e.g., a true “lights out” factory), the
spective of the Network Lens and the Lean tools used to reality is that all business systems involve people. To
address complexity in systems is that much of the aca- account for this, we require the fourth and most expan-
demic literature related to bottleneck and queueing anal- sive perspective on Lean which we term the Organiza-
ysis assumes stationarity. That is, over the long run tion Lens.
resource utilizations are stable, which implies that a Some of the Lean practice literature has recognized
unique bottleneck (or at least a small set of bottlenecks the need for an organizational perspective. For example,
and near bottlenecks) will govern system behavior. But in Spear and Bowen (1999) noted that a Lean culture
practice, it is possible, even likely, that neither demand requires attention to training, communication and other
nor capacity are stationary. Demand may be predictably human behaviors within the organization. Liker (2004)
HOPP AND SPEARMAN 11

emphasized developing people and organizational learn- day will notice and remember the days that confirm
ing in his 14 principles of the TPS. While these are valid her belief more than the days that refute it.
observations about the culture at Toyota, they do not • Loss Aversion—The motive to avoid losses is stronger
leverage modern behavioral science. In order make full than the motive to achieve gain. For example, a bakery
use of the Organization Lens we need to incorporate owner may be more upset by a day on which many
human behavior more scientifically. loaves of bread are thrown away than a day on which
A few authors (e.g., de Treville & Antonaki, 2006; bread ran out by mid-afternoon. The former is a mea-
Hopp, 2018; Parker, 2003) have recognized that the litera- surable loss, since expensive ingredients were wasted.
ture on the psychology of work is relevant to Lean The latter is a foregone gain (the bakery could have
because it is ultimately implemented by people with jobs. sold more bread and made more profit). If the owner
Designing those jobs to make work rewarding and moti- over emphasizes the loss and under emphasizes the
vational can have a significant influence over the success gain, she will order less than the profit maximizing
of Lean implementations. But, so far, operations scholars amount of bread. Indeed, experimental evidence has
have made only limited use of the insights from shown that people have a tendency to do precisely this,
psychology. provided the profit margin is sufficiently high
A more extensive, and largely untapped, resource is (Schweitzer & Cachon, 2000).9
the wide array of cognitive research into heuristics and
biases that has been developed by behavioral and deci- If the hindsight bias extended by confirmation bias
sion scientists since the 1970s. Kahneman (2011) uses the leads the bakery to make too much bread it will result
concepts of “System 1” and “System 2” to describe this overproduction waste. If the loss aversion bias leads the
body of work. System 1 (“fast thinking”) refers to the bakery to make too little bread it will lead to the waste of
involuntary, automatic, effortless responses our brains excess waiting on the part of customers who must come
make to the world around us.8 System 2 (“slow think- back another day if they want to get bread. The time
ing”) refers to the methodical, logical and energy inten- buffer represented by this waiting could be particularly
sive responses we make to selected situations. Because costly, because it may frustrate customers and decrease
each of us makes some 35,000 decisions every day, it is their likelihood of making future purchases. Either way,
impossible for us to use System 2 thinking for more than the result will be a wasteful departure from the goal of
a tiny fraction of our choices. Fortunately, because it has Lean.10
been shaped by evolutionary forces, System 1 usually A possible implication from the research into psycho-
works well. But, because our world differs from the one logical biases is that people need more training in proba-
in which it evolved or for other reasons, System 1 heuris- bilistic thinking to deal with problems involving
tics sometimes lead to unconscious bias or other irratio- uncertainty. Another conclusion could be that planning
nal decisions. under uncertainty in many situations should not be done
Recent research shows that the occasional irrational- by intuitive feel at all, but instead should make use of a
ity from System 1 thinking is often predictable (see for data-based decision support system. A small business like
example, Ariely, 2008). A few common biases that affect a bakery has neither statistically trained personnel nor a
our everyday thinking and almost certainly come into decision support system. As a result, they presumably
play when people make decisions about the design and make errors like the one described above routinely. The
operation of Lean systems include: “newsvendor” scenario presented by the bread planning
example is a very simple case where biases lead to waste.
• Hindsight Bias—We judge decisions by outcome rather Large organizations using sophisticated ERP systems
than soundness of method. For example, suppose a presumably have the computational power to address the
bakery owner is clueless about how much bread to newsvendor aspect of their inventory planning problems.
bake on the day before Thanksgiving and uses guess- But even they often use highly visible sales to represent
work and poor logic to justify making a very large demand, while omitting the invisible lost sales. As a
amount of bread. If she gets lucky with an unusually result, decisions such as inventory planning, which are
high demand day, she is apt to say “See! I knew it all subject to uncertainty, may be compromised by decision
along” instead of recalling the uncertainty and contro- biases in large, as well as small, firms.
versy surrounding her decision. Subtler situations related to Lean implementations
• Confirmation Bias—We look for evidence that con- where incentives and/or decisions may also be distorted
firms our beliefs. For example, a bakery owner who, by similar biases include choosing training activities to
because of a past lucky decision or another reason, carry out, choices made during kaizen events, and con-
feels the bakery should be making more bread each clusions drawn from Gemba walks. In each of these, we
12 HOPP AND SPEARMAN

are prone to focus on outcomes rather than processes, see When asked what percentage of customers wanted their
evidence that we want to see, and inconsistently empha- bread sliced, the staff could not say with any precision
size avoiding losses at the expense of passing up gains. but agreed that it was well over half. Nevertheless, it took
We see examples of faulty decision making in Lean a good deal of cajoling and reminding to get the staff to
implementations all the time. However, as is always pos- agree to slice half of the bread bagging as an experiment.
sible when uncertainty is involved, sometimes bad When it quickly became apparent that this policy never
(i.e., not based on logical consideration of the informa- resulted in a stockout of unsliced bread but almost
tion available) decisions lead to good outcomes. When always required un-bagging and slicing some of the
this occurs, the two of us are fond of shaking our heads unsliced bread, the staff recognized on their own the
and saying “oh well, it's better to be lucky than smart.” need to increase the proportion of bread sliced before
But, while this is tautologically true if lucky means get- bagging. They soon converged on a policy of slicing 80%
ting good results and smart means making well-reasoned of the bread and agreed that the new system was clearly
decisions, it may not be true in real life when we consider better for them and for the customers.11
cognitive biases and their impacts on future decisions. As The point of this example is two-fold. First, even in
Kahneman points out, even a single observation can cre- very simple execution-level activities, uncertainty can
ate an associative memory that System 1 can mistake for present decision problems for which our System 1 think-
a pattern. This plus confirmation bias and hindsight bias ing processes are ill-equipped and prone to errors. Sec-
can result in a lucky outcome distorting decision making ond, implementing Lean improvements requires more
for a long time. For example, if the bakery owner makes than laying them out in logical fashion for people to
a bad decision and bakes too much bread but gets lucky adopt. People need to be engaged in the implementation
and sells it, the associative memory reinforced by confir- in a way that engages their System 2 thinking and over-
mation bias could lead to many more bad decisions to comes their System 1 biases. This may be why the Lean
overbake in the future. literature is replete with claims (Durin, 2018; Liker &
These and other biases can certainly affect manage- Rother, 2011) that implementation failures are due to fac-
ment decision making related to Lean, such as choosing tors such as lack of involvement by top management,
production quantities, incentives, or motivational activi- unclear goals, or inadequate training. What these plausi-
ties. But they can also affect behavior in the execution of ble but ill-defined reasons may really be saying is that the
Lean. An illustrative example occurred in the bread- Lean implementations failed to take into account the
slicing operation of a small family-owned bakery. The way real people think and behave. It may also be behind
basic bread production process is to bake, cool, slice, bag, the success of bottom-up problem-solving processes like
and sell the bread. However, because some customers Kaizen (Shingo, 2007) that engage those responsible for
prefer their bread unsliced and there is uncertainty about implementing Lean policies in the design of them. How-
the proportion of customers who will want unsliced ever, we need to be careful about putting too much faith
bread on any given day, there is an inherent challenge in in the simple act of involving workers in designing
matching bread to customer preferences. In this particu- improvements. We are all prone to cognitive biases that
lar bakery, this challenge had been addressed for as long can blind us to effective alternatives. Therefore, how we
as anyone could remember by bagging unsliced bread involve people in the search for ways to implement Lean,
once it was cool and then un-bagging and slicing it for what data are provided, how questions are posed, and
any customer who wanted sliced bread. This resulted in how well people are prepared to think about problems
extra work for the staff to unbag and re-bag the bread that involve uncertainty, all matter in the effectiveness of
(processing waste), as well as delay while the customer bottom-up problem solving.
waited for the bread to be sliced (waiting waste). More- Psychological biases may even influence the effective-
over, because it required slicing one loaf at a time, rather ness of continual improvement processes like Kaizen. For
than slicing all the loaves in a repetitive batch mode, it example, the bias toward loss aversion can lead people to
involved even more processing waste due to the loss of avoid setting high-achieving goals. An unachieved goal is
an economy of scale. a loss, while exceeding a goal is a gain. Hence, a low goal
Recognizing this, the new store owner suggested slic- makes it easier to avoid a painful loss, albeit at the
ing at least some of the bread before bagging it. This led expense of a reduced likelihood of the satisfaction of a
to a surprisingly difficult discussion with the staff. Some high-level outcome. Moreover, once a goal is set, people
of them focused on specific customers who strongly pre- have more incentive to reach it than to exceed
ferred unsliced bread (What if Mrs. Smith cannot get her it. Kahneman (2011) explains that in many cases, people
unsliced loaf!). Others agreed that slicing some of the will reduce their efforts once they have reached a specific
bread made sense but could not agree on how much. goal because they see no point going above and beyond.
HOPP AND SPEARMAN 13

The loss aversion bias makes us more concerned about (e.g., by reorganizing into modular layouts. Most firms
avoiding losses than about striving to achieve gains. never get to the Organization Lens. This is probably not
Finally, underappreciating the nature of System 1 and too surprising given that the Lean literature is devoted
the biases it introduces can lead management to conclude primarily to Process Lens descriptions, with some cover-
that implementing a new policy should be simpler than it age from the Flow and Network Lens perspectives and
actually is. Managers often assume that if the logic of a almost entirely lacking in an Organization Lens focus.
new way of doing things is manifest then this should be In contrast, Toyota adopted a Network Lens focus on
enough to get people to adopt it (e.g., how could anyone flow simplification very early on. In the assembly plants
not appreciate the efficiency benefits of slicing bread using the 2-shift schedule, this led to mixed-model pro-
before bagging to avoid unnecessary un-bagging and re- duction to simplify flow in final assembly and cellular
bagging?). But this neglects the inertia created by associa- manufacturing in component lines (see Hall, 1983;
tive memory and confirmation bias. As in the case of the Schonberger, 1982; Shingo, 1985 for a description of these
bread slicing and bagging example, an effective Lean and other TPS methods). Spear and Bowen (1999)
implementation plan needs to be designed in a way that described this emphasis on reducing complexity in flows
encourages staff to engage their System 2 thinking pro- as the DNA of the TPS. The first three of the four rules
cesses and overcome their unconscious biases. they used to encapsulate TPS were:
Because little has been done to make use of cognitive
science in academic studies of Lean, and almost none of Rule 1 – All work shall be highly specified as
the insights from behavioral science have been incorpo- to content, sequence, timing, and outcome.
rated into Lean implementations in industry, there is a Rule 2 – Every customer-supplier connection
huge opportunity to leverage the insights about psycho- must be direct, and there must be an unam-
logical biases in both Lean research and practice. biguous yes-or-no way to send requests and
receive responses.
Rule 3 – The pathway for every product and
6 | LINKING THE LENSES service must be simple and direct. (Spear &
Bowen, 1999)
Each of the above lenses can serve as a research perspec-
tive on Lean and can lead to insights that help guide Rules 1 and 2 are prerequisites to the simple flows called
Lean practice. But fully embracing Lean as part of a com- for in Rule 3. Note that Spear and Bowen gave this Net-
prehensive change management program requires more work Lens perspective primacy over the Process Lens
than making use of individual lenses. It also requires emphasis on waste reduction and Toyota tools (Kanban
sequencing and integrating the perspectives of the lenses and 5S).
in a way that facilitates continual improvement. To illus- Toyota also adopted an early Flow Lens focus on vari-
trate we invoke the famous “two-shift” practice of ability reduction. In particular, they made clever use of
Toyota. their variability-buffering strategy to make variability vis-
In the 1980s when almost all auto manufacturers ran ible and only then made use of Process Lens methods to
their assembly plants on a three-shift basis, Toyota made eliminate the variability and the waste it causes. The two-
use of a two-shift system consisting of two 8-hr shifts per shift strategy was essential to this stage of Toyota's Lean
day separated by two 4-hr preventive (PM) maintenance evolution because, in addition to providing time for
periods. On the surface, it appeared that Toyota was maintenance, the PM periods could be used for overtime
sacrificing 8 hr/day of capacity in very expensive facili- if needed to meet the daily production quotas. As such,
ties. However, when viewed through the lenses of Lean, the PM periods represented a very large capacity buffer
it becomes clear that they were doing something much against variability in the production or demand rate. This
more profound. permitted Toyota to reduce inventory (raw materials,
A key difference between Toyota's practices and those work-in-process, finished goods) buffers without inflating
of their contemporaries and many firms implementing time buffers (delays in meeting customer demands).
Lean today is the sequence in which they employed the Reduced inventory helped to identify sources of vari-
lenses. Other firms have tended to start with a Process ability in the production system, as Toyota described with
Lens focus on obvious waste (e.g., with 5S reorganiza- the well-known analogy of lowering the water in a river
tions of individual processes), progress to a Flow Lens to uncover the rocks. In less poetic terms, without excess
focus on simple flows (e.g., by using Value Stream Map- WIP in the system, each glitch in the system would be
ping to identify inefficiencies), and finally adopt a Net- immediately apparent because of the disruption in flow it
work Lens focus on the overall production system would cause. Toyota further highlighted glitches with an
14 HOPP AND SPEARMAN

Andon system that allowed workers to stop the line customer preferences, technology, and many other
whenever they encountered a problem either with the factors, it is impossible to permanently optimize a pro-
quality of the product or with the procedure being used. duction system. A continually evolving and adapting sys-
To torture the water analogy a bit, the Andon system was tem is needed. To achieve this, it is essential to punctuate
analogous to a lookout on the ship watching for rocks, ongoing Process Lens vigilance with periodic reviews of
while the available PM periods were analogous to having the bigger pictures provided by the Flow and Network
the capability to dig channels around the rocks so that Lenses.
they could be uncovered without harming the ship (fac- An example of this type of iteration was on display
tory) or passengers (customers). when one of us took a sabbatical at Motorola. The com-
With this systematic process for making problems vis- pany had recently announced completion of their cele-
ible in place, all of the Process Lens techniques become brated Six Sigma initiative and had launched a “10X
more effective. Without it, efforts at waste reduction can Cycle Time” initiative with a goal to radically reduce
be “for show rather than dough.” That is, as we have seen (by a factor of 10) their design, production, and fulfill-
many times, Lean initiatives can lead to inventory reduc- ment cycle times. This ambitious goal forced the firm to
tions that have little (or even negative) impact on costs, eliminate process stages, and even entire processes.
capacity enhancements that affect only non-bottlenecks, When this became apparent, a Vice President remarked
or internal flow enhancements that do not improve cus- that he wished they had done 10X Cycle Time before Six
tomer service, all of which do not serve the core strategy Sigma, because they could have avoided a great deal of
of the organization. Making problems visible helps waste (Process Lens) improvement work on process that were
seekers find the strategically important waste. The fact removed through the (Network Lens) review of the 10X
that Toyota premised their direct waste elimination program. But he may have been holding himself to too
efforts with production-network rationalization and flow- high a standard. Anticipating the need for speed at such
variability reduction phases may be an important reason a detailed level a decade earlier would have required
they were able to practice Lean more effectively than remarkable foresight. In a changing world iteration will
most of their rivals for decades. Furthermore, in the early always be necessary.
2000s, after decades of variability reduction, Toyota evi- We can summarize the insights from this discussion
dently concluded they no longer needed large capacity captured in Figure 2 as:
buffers to protect customer deliveries and began
switching assembly plants to three-shift schedules. 1. Think big before thinking small. A system review that
Finally, although neither Toyota nor anyone else has simplifies the production network and a flow focus
fully exploited the behavioral focus of the Organization that identifies bottlenecks helps to focus waste elimi-
Lens, Toyota has always devoted time to the culture of nation efforts on processes that matter to overall
Lean.12 For example, they use kaizen as a way to gain performance.
worker buy-in to the continual improvement process and 2. Adjust buffers to facilitate exploration and exploitation.
to inject a worker perspective into any job-design deci- Optimizing variability buffers is a vital part of Lean
sions made as part of the Lean transformation. implementation. But buffers can also be adjusted
Figure 2 provides a high-level schematic of the stag- to reveal the sources of variability. This is done by
ing of lenses in the Toyota scenario. Note that we have reducing WIP buffers, so that disruptions affect
included reverse arrows to allow for repeated cycling flow quickly, while increasing other buffers
through the various perspectives. The reason for this is (e.g., capacity or finished goods) to protect cus-
that, given persistent change in product volumes, mix, tomer service.

ORGANZATIONAL LENS Organizational Transformation

NETWORK LENS FLOW LENS PROCESS LENS

Rationalize/Simplify Reveal/Reduce Eliminate Process
Flow Network Flow Variability Level Waste

FIGURE 2 Staged use of lean lenses

in practice
HOPP AND SPEARMAN 15

3. Pursue physics and psychology in parallel. Each step in variability to describe the causes of waste in a flow. The
the Lean process must be executed by people and each Network Lens makes use of a network representation of
change that is made affects someone's job. It is there- a production or service system, and particularly the con-
fore essential to think about both the physics of flows cept of bottlenecks, to understand causes of waste and to
(variability, bottlenecks, etc.) and the psychology of identify areas of maximum leverage in complex systems
work (job design, motivation, cognitive bias, etc.) at of interconnected flows. The Organization Lens draws on
each stage of Lean implementation. behavioral science to anticipate human reactions to Lean
4. Iterate, iterate, iterate. Although Lean as a title may policies and to guide designs that work with human ten-
have an expiration date, the pursuit of efficiency is a dencies rather than against them.
never-ending journey. Production systems must be Taken together, the four Lenses offer a perspective on
adapted constantly to changes in the market and pro- Lean research that leverages rapidly advancing fields like
duction environments and organizational cultures Data Analytics, Network Science, and Cognitive Psychol-
must evolve to support and accommodate these ogy, and extends the range of questions amenable to
changes. empirical and analytical studies. The Lenses also offer a
framework for enhancing Lean management systems by
Although the schematic in Figure 2 is meant primar- making of the insights from these fields, and by
ily as a practice guide, it can also serve as a basis for iden- balancing implementation efforts across the levels of the
tifying research opportunities. For example, Lean enterprise from process mechanics to organizational cul-
scholars could make empirical studies of the effectiveness ture. As such, these Lenses can help both scholars and
of different activity sequences in Lean implementations practitioners see the opportunities of Lean more clearly
to provide clearer guidance to firms adopting Lean. They and promote the pursuit of efficiency more effectively.
could study the impact of WIP level on organizational
learning to see whether the approach used by Toyota ORCID
with their two-shift system is broadly effective. They Wallace J. Hopp https://orcid.org/0000-0003-1586-8606
could examine the impact of cognitive bias on specific
Lean practices as a prelude to finding better ways to coor- ENDN OTE S
dinate the physical and psychological sides of Lean. 1
The separation of OM sub-fields by strategic focus does not mean
Many other research opportunities are likely to arise they do not overlap. For example, a central tool of efficiency man-
from examining Lean through the combined perspectives agement is variability reduction and an important means of vari-
of the four Lenses. ability reduction is quality management. Hence, efficiency and
quality management are often pursued in concert, as evidenced
by the blurring of the two under the popular heading of Lean Six
Sigma.
7 | C ON C L U S I ON S 2
Although this buffering framework was introduced in the 1990's
as part of the Factory Physics framework, Taiichi Ohno, the father
Explicitly addressing Lean progressively through the Pro- of the Toyota Production System, recognized the central role of
cess, Flow, Network, and Organization Lenses can help variability as a cause of waste much earlier in his tri-dimensional
practitioners from getting stuck at the lowest level of description of waste as muda (waste), mura (variability) and muri
Lean execution. It is unfortunate that so much press has (overburden) (see Hopp, 2018 for a discussion).
been given to the 7 types of waste and specific practices 3
If you asked us to give the single most useful definition of Lean
such as Kanban and 5S. Neither categories of waste nor this would be it. By leveraging the most powerful insights of Fac-
individual practices lead to an understanding of the tory Physics into the nature and causes of waste, it provides a
underlying causes of waste. The “5 Whys” approach from robust modeling framework for research and a practical diagnos-
tic tool for practice. However, tempered by the mantra “all
Toyota seeks to expose the primary causes of waste, as do
models are wrong, but some are useful”, we recognize that alter-
more formal techniques like fault-tree analysis, root-
nate definitions can be useful too. Therefore, we present this def-
cause analysis and fishbone diagramming. But all of inition as only one among four, rather than as the Holy Grail
these are generic methods, which depend on a knowledge of Lean.
of the behavior of flows on the part of the user to be effec- 4
Note that variability propagates across measures. For instance,
tive in diagnosing waste in production and service variability in quality impacts variability in the time and quantity
systems. of production by impacting yield and rework. Variability in the
To provide explicit guidance for Lean implementation time and quantity of output results in variability of costs and reve-
we need to go beyond classification and practices implied nues. Consequently, variability is an important connection
by the Process Lens. This is what the Lenses introduced between operational and financial metrics.
in this paper do. The Flow Lens leverages the science of
16 HOPP AND SPEARMAN

5 12
Many of the quoted lead times would also be 4 weeks or less and Of course, Toyota does not use the generic term Lean. Instead,
hence, unlike the fixed 10-week lead times, would be competitive their culture building initiatives are made under the eponymous
with those of the rival firm. Toyota Production System banner. It is not clear that Lean,
6
A clever example of creating the institutional readiness to whether defined broadly as efficiency management or more nar-
implement an emergency response is the “chaining” structure rowly as some subset of the efficiency space, can serve as a moti-
proposed by Jordan and Graves (1995). By equipping factories vational title. It may be more effective in practice for firms to
with overlapping capabilities, this system enables adjustment of create their own titles (“X Production System” or “Y Business
product-to-plant assignments to accommodate wide fluctuations System”) to describe a collection of Operations Management
in demand mix. Although few would describe this as a Lean ini- ideas and practices (from the Lean/efficiency area, as well as
tiative, it facilitates meeting demand with less installed capacity, other OM subfieklds) combined with insights from other man-
which is clearly an efficiency measure. So, in the terminology of agement fields (e.g., strategy, marketing, organizational behav-
this paper, it is a means to Lean. ior), and firm-specific goals and practices. Whether firm-specific
7 titles are more motivational than generic terms is an open
Another example of a contingency plan is the tailored base-surge
research question.
policy proposed by Allon and Van Mieghem (2010) in which an
efficient supplier is used to provide a constant supply, while a
responsive (but more expensive) supplier is used to meet surges in RE FER EN CES
demand. de Treville, Cattani, and Saarinen (2017) address the
Allon, G., & Van Mieghem, J. (2010). Global dual sourcing: Tailored
challenge of the responsive supplier by describing how producing
base-surge allocation to near- and offshore production. Man-
a portfolio of time-sensitive and time-insensitive products can
agement Science, 56(1), 110–124.
make effective use of capacity and allow competitive production
Ariely, D. (2008). Predictably irrational. New York, NY:
in a high-cost environment. Bakeries use a version of the tailored
HarperCollins.
base-surge approach for commonly available items (milk, eggs,
Belekoukiasa, I., Garza-Reyesb, J., & Kumarc, V. (2014). The impact
chocolate chips) with a wholesale supplier as the efficient supplier
of lean methods and tools on the operational performance of
and a grocery store as the responsive supplier.
manufacturing organisations. International Journal of Produc-
8
Gigerenzer and Goldstein (1996) coined the more descriptive “fast tion Research, 52(18), 5346–5366.
and frugal” label for System 1 thinking in order to emphasize that Browning, T., & Heath, R. (2009). Reconceptualizing the effects of
it relies on heuristics that limit both information search and lean on production costs with evidence from the F-22 program.
computation. Journal of Operations Management, 27(1), 23–44.
9 de Treville, S., & Antonaki, J. (2006). Could lean production job
Experiments indicate that people tend to over-order in settings
like that of the bread baking scenario when the profit margin is design be intrinsically motivating? Contextual, configurational,
low, which was originally thought to violate the behavior and levels-of-analysis issues. Journal of Operations Manage-
predicted by prospect theory. However, Long and Nasiry (2015) ment, 24(2), 99–123.
pointed out that if the decision maker defines gains and losses rel- de Treville, S., Cattani, K., & Saarinen, L. (2017). Technical note:
ative to an aspiration level other than the status, as Kahneman Option-based costing and the volatility portfolio. Journal of
and Tversky (1979) allowed for, then the observed behavior is Operations Management, 49-51, 77–81.
consistent with prospect theory. Durin, M. 2018. Why most companies are failing with Lean imple-
10 mentation? Kaizen Institute Blog. Retrieved from https://www.
If bread seems to mundane an example of how cognitive biases
kaizen.com/blog/post/2018/07/05/why-are-most-companies-
can distort operational decisions, consider the work of Gray,
failing-with-lean-implementation.html
Esenduran, Rungtusanatham, and Skowronski (2017), which
Ford, H. (1922). My life and work. New York, NY: Doubleday,
studied companies that offshored their production to reduce cost
Page & Co.
only to reverse their decisions and bring production back home
Gigerenzer, G., & Goldstein, D. (1996). Reasoning the fast and fru-
at great expense. By scrutinizing these decisions in light of the
gal way: Models of bounded rationality. Psychological Review,
research on heuristic decision making, the authors concluded
103, 650–669.
that the firms fell prey to well-known decision biases that led
Gilbreth, F. (1911). Motion study. New York, NY: D. Van
them to rely on an overly simplistic “lowest per-unit landed-cost”
Nostrand Co.
that omitted important cost considerations such as quality issues
Goldratt, E., & Cox, J. (1986). The goal: A process of ongoing
and intellectual property loss.
11
improvement. New York, NY: North River Press.
The cognoscenti are likely to point out that finding an optimal Gokpinar, B., Hopp, W., & Iravani, S. (2010). The impact of
percentage of bread to pre-slice is a Newsvendor problem that misalignment of organization structure and product
requires the (underage) un-bagging and re-bagging cost to slice architecture on quality in complex product development. Man-
and unsliced loaf, the (overage) cost of a disappointed customer agement Science, 56(3), 468–484.
who wants an unsliced loaf but must accept a sliced loaf or go Gray, J., Esenduran, G., Rungtusanatham, M., & Skowronski, K.
without, and the distribution of the demand for unsliced bread. (2017). Why in the world did they reshore? Examining small to
But this is probably too much System 2 thinking to devote to this medium-sized manufacturer decisions. Journal of Operations
problem when there are bigger problems (e.g., determining bake Management, 49-51, 37–51.
quantities and optimizing the advertising strategy) that can gen- Hall, R. (1983). Zero inventories. Homewood, IL: Dow Jones-Irwin.
erate more value from analytic solutions.
HOPP AND SPEARMAN 17

Hopp, W. (2011). Supply chain science. Long Grove, IL: Waveland Shingo, S. (1985). A revolution in manufacturing: The SMED system.
Press. Portland, OR: Productivity Press. (Original Japanese edition,
Hopp, W. (2018). Positive lean: Merging the science of efficiency Shinguru Dandori, published by the Japan Management Associ-
with the psychology of work. International Journal of Produc- ation, Tokyo, 1978).
tion Research, 56(1–2), 398–413. Shingo, S. (2007). Kaizen and the art of creative thinking. Belling-
Hopp, W., & Spearman, M. (2004). To pull or not to pull: What is ham, WA: Enna Products Incorporated. (Original Japanese edi-
the question? Manufacturing & Service Operations Management, tion published by Shobo Publishing Company, Tokyo, 1959).
6(2), 133–148. Spear, S., & Bowen, H. (1999). Decoding the DNA of the Toyota
Hopp, W., & Spearman, M. (2008). Factory physics (3rd ed.). Long production system. Harvard Business Review, 77(5), 96–106.
Grove, IL: Waveland Press. Spearman, M. L., & Hopp, W. J. (2019). Fundamentals of
Jordan, W., & Graves, S. (1995). Principles on the benefits of operations science. Working paper, University of Michigan, Ross
manufacturing process flexibility. Management Science, 41(4), School of Business.
577–594. Sugimori, Y., Kusunoki, K., Cho, F., & Uchikawa, S. (1977). Toyota
Kahneman, D. (2011). Thinking, fast and slow. London, England: production system and kanban system materialization of just-
Penguin Books. in-time and respect-for-human system. International Journal of
Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis Production Research, 15(6), 553–564.
of decision under risk. Econometrica, 47(2), 263–292. Suri, R. (1998). Quick response manufacturing: A companywide
Krafcik, J. (1988). Triumph of the lean production system. Sloan approach to reducing Lead times. Portland, OR: Productivity
Management Review, 30(1), 41–52. Press.
Lean Enterprise Institute. (2019). What is lean? Retrieved from Taylor, F. W. (1911). The principles of scientific management.
https://www.lean.org/WhatsLean/ New York, NY; London, England: Harper & Brothers.
Liker, J. (2004). The Toyota way: 14 management principles from the Ulhassan, W., von Thiele Schwarz, U., Thor, J., & Westerlund, H.
world's greatest manufacturer. New York, NY: McGraw-Hill. (2014). Interactions between lean management and the psycho-
Liker, J., & Rother, M. (2011). Why lean programs fail. Lean Enter- social work environment in a hospital setting – A multi-method
prise Institute. Retrieved from https://www.lean.org/Search/ study. BMC Health Services Research, 14, 480–490.
Documents/352.pdf Vidal, M. (2007). Lean production, worker empowerment, and job
Long, X., & Nasiry, J. (2015). Prospect theory explains newsvendor satisfaction: A qualitative analysis and critique. Critical Sociol-
behavior: The role of reference points. Management Science, 61 ogy, 33, 247–278.
(12), 3009–3012. Womack, J., & Jones, D. (1996). Lean thinking: Banish waste and
Martinez, D. A., Kane, E. M., Jalalpour, M., Scheulen, J., create wealth in your corporation. New York, NY: Simon &
Rupani, H., Toteja, R., … Levin, S. R. (2018). An electronic Schuster.
dashboard to monitor patient flow at the Johns Hopkins Hospi- Womack, J., Jones, D., & Roos, D. (1991). The machine that
tal: Communication of key performance indicators using the changed the world: The story of lean production. New York, NY:
Donabedian model. Journal of Medical Systems, 42(8), 133. Harper Perennial.
Moraro, J., Lemstra, M., & Nwankwo, C. (2016). Lean interventions Yin, Y., Stecke, K., Swink, M., & Kaku, I. (2017). Lessons from seru
in healthcare: Do they actually work? A systematic literature production on manufacturing competitively in a high cost envi-
review. International Journal for Quality in Health Care, 28(2), ronment. Journal of Operations Management, 49-51, 67–76.
150–165. Zarbo, R., & D'Angelo, R. (2006). Transforming to a quality culture:
Lopes Negr~ao, L., Godinho Filho, M., & Marodin, G. (2017). Lean The Henry Ford Production System. American Journal of Clini-
practices and their effect on performance: A literature review. cal Pathology, 126(Suppl 1), S21–S29.
Production Planning & Control, 28(1), 33–56. Zipkin, P. (2000). Foundations of inventory management. New York,
Parker, S. (2003). Longitudinal effects of lean production on NY: McGraw-Hill.
employee outcomes and the mediating role of work characteris-
tics. Journal of Applied Psychology, 88(4), 620–634.
Schmenner, R., & Swink, M. (1998). On theory in operations man-
agement. Journal of Operations Management, 17, 97–113. How to cite this article: Hopp WJ,
Schonberger, R. (1982). Japanese manufacturing techniques: Nine Spearman MS. The lenses of lean: Visioning the
hidden lessons in simplicity. New York, NY: Free Press. science and practice of efficiency. J Oper Manag.
Schweitzer, M., & Cachon, G. (2000). Decision bias in the 2020;1–17. https://doi.org/10.1002/joom.1115
newsvendor problem with a known demand distribution:
Experimental evidence. Management Science, 46(3), 404–420.