Aryma Labs

How Google's announcement to not deprecate third party cookies impacted us positively 🚀. In information theory, there is a concept called Entropy. If you want a good refresher, I would highly recommend the article by Naoki (Link in comments). Basically, Low Entropy means receiving very predictable information. High entropy means receiving very unpredictable information (some also define it as an element of surprise). Earlier, there was very high entropy in the marketing measurement and attribution space because Google kept dilly dallying over cookie deprecation. Nobody was exactly sure when it would happen. And even if they went through with it, people had a lot of questions over their privacy sandbox. Would it be accurate? How trustworthy it would be ? etc. were the questions posed by the marketers and advertisers. But On July 22nd,2024 the high entropy turned it to low entropy. Meaning a lot of uncertainty was removed. Google announced that it won't deprecate 3rd party cookies but rather give the option directly to the users to opt-in or not (much like Apple's ATT solution). This signal was enough for marketers and advertisers to really double down on privacy proof marketing measurement methods for 2025 and beyond. We at Aryma Labs saw a significant jump in the number of calls for MMM services/ demos in days just after the Google announcement. But we wanted to be more scientific about this proclamation. 📌 The ITSA method We have been implementing ITSA (Interrupted Time Series Analysis) method for our clients to help them understand the effect of special campaign launches or product launches on the KPI. We decided to apply ITSA to our own data set. This data set comprises of number of calls for MMM services/Demo requests from our website starting from Jan 1st to Aug 16th of this year. Overall out of 8 months, there were 32 days where we got at least 1 service / demo request (pls refer the image). But the most interesting thing to observe is that there has been an increase in flurry of activities post July 22nd. The Highlights of ITSA analysis are as follows (we will be publishing a detailed research paper on this soon). ▪ On an average, before intervention (July 22nd event), we had approximately 1 service request/demo request. ▪ Post intervention, we see that we are getting additional 1 more request !! We also ran a BSTS model to compare, and found similar results as ITSA. The estimate of causal effect of intervention is 0.91. An additional approx. 1 request on average is observed in the post-intervention period compared to the pre-period. We are sure we are not the only ones to experience this uplift in service/demo requests for MMM. Our rivals too could be experiencing the same. As they say, the boat rises with the tide. The tide of privacy proof marketing measurement and attribution is on the rise.

I am Excited to share the complete case study on our recent versatile MMM project - Modeling Crowd Attendance through MMM !! Earlier last week, I had briefly talked about this case study and how we at Aryma Labs not only excel at OG MMM (CPG/FMCG) but also with implementation of MMM in other domains like Sports, QSR, Finance, Gaming, E-commerce and Pharma. I am really proud of the team for the work done on this project as it required out of the box thinking not only statistically but also from business problem stand point. Kudos Tannishtha Sen Febin P Babu Soham Giri and Venkat Raman. Also thanks Soham (Our in house Football Aficionado ⚽) for the beautiful work on the case study.

What should your Marketing Measurement Solutions Mix for 2025 look like ? With less than 4 months left in 2024, many companies have already started planning Marketing budgets for 2025. In my conversations with CEOs, CMOs and Brand Managers, one recent topic that keeps springing up is - what should the Marketing Measurement Solution portfolio look like in 2025? In my previous posts (link in comments), I had covered how Google's ATT like opt-in in 2025 will create huge marketing measurement gaps. One casualty to Google's opt-in will be Multi Touch Attribution (MTA). MTA will become less reliable. Hence any investment into MTA going forward will be depreciative in nature. We are strongly advising companies to reduce the budget on MTA or if possible totally stop spending on MTA. 📌 Which Marketing Solutions to invest in 2025? We would strongly recommend having a trifecta approach. Invest more in Marketing Mix Modeling (MMM) capabilities supported by Experimentation and Causal Experiments initiatives. Since early last year, we at Aryma Labs had developed a philosophy of triangulation (MMM, Experimentation and Causal Inference) for Marketing Measurement and Attribution. We believe this will be a very comprehensive and accurate Marketing Measurement solution combination going into 2025. Predominantly Budget optimization in MMM is done for marketing channels. But we developed an interesting simulation where we tried to do budget optimization between Marketing Measurement solutions. As with any budget optimization, the methodology is that money is taken out from low ROI mediums and put into high ROI mediums. We believe MTA will be a low ROI solution in 2025 and hence suggest to take away spends from it. 📌 Where should that money go? We suggest more deployment on MMM. Followed by Experimentation (A/B testing, Geo Testing) and Causal Experimentation like DID. Note: The numbers in the chart are not cost for 1 MTA or 1 MMM model. Rather it is the cumulative spend on whole MMM, MTA or Experimentation solution which could comprise many experiments/models.

Hello Everyone, Excited to announce the official launch of the MMMGPT🚀 Interest in Marketing Mix Modeling (MMM) is on the rise due to the growing need for a privacy-proof marketing measurement solution. Many are eager to learn and experiment with MMM. Ridhima Kumar and I often receive messages from around the globe requesting resources on MMM. We could have put all our knowledge into a book. But we did the next best and perhaps the futuristic thing. We created world's first MMMGPT !! MMMGPT is a RAG based model and not a simple wrapper on ChatGPT. It took us many long days to get the algorithm right and most importantly get the prompts right. In the future I will perhaps do a detailed technical video explaining our RAG architecture and the innovative pieces that we had to code up to make our MMMGPT extremely accurate. But for now, in this video I provide a detailed demo of our MMMGPT. We are also simultaneously launching MMMGPT Enterprise (Refer the details in the MMMGPT website). If you're looking to enhance your knowledge on MMM and access the most accurate insights, give MMMGPT a try. Link to sign up in the comments. 

Impact of Google's ATT like Opt-in on various Marketing Measurement Solutions If you are investing heavily on Multi Touch Attribution (MTA), now is the time to cut down on it and if possible totally cease spending on MTA solutions. Why? Ans: Google's ATT like opt-in (to be implemented from 2025). When Apple launched its App Tracking Transparency (ATT) in 2021, its opt-in rates initially were around 5-10% according to various reports. By various conservative estimates, Google's opt in rates will be around 5-10% as well. MTA historically has had a lot of flaws (e.g. incomplete data, non statistical methodology). But now the problem for MTA will be exacerbated as Google's shift to an opt-in model will result in fewer users consenting to being tracked, leading to larger gaps in the user journey data that MTA depends on. 📌 So what should Marketers do? In one simple sentence - Adopt MMM and Experimentation (more specifically causal experimentation). MMM and Experimentation are going to be less or totally unaffected by Google's opt-in. This makes it a must have solution for marketers going forward. We highly encourage our clients and prospective clients to invest more in MMM and Experimentation. 📌 What about the Granularity problem of MMM? Yes, one key advantage that MTA always had over MMM was its ability to provide granular insights (which campaigns worked, which TV channels worked etc.). But not anymore. Earlier last year, we made a remarkable breakthrough. We devised a statistical technique called GTA-F (based on game theory and information theory). This patent pending technique not only solves the granularity problem but also helps validate attribution of in-platform metrics. Link to the detailed case study in comments. So granularity is no longer the Achilles' heel of MMM. In summary - Adopting MMM and Experimentation offers a comprehensive view of your marketing effectiveness in light of the evolving measurement landscape. P.S : Tony Evans thanks for the excellent idea on the chart.

We are Growing Rapidly 🚀 . We are hiring !!

One of the things that I am really proud of at Aryma Labs is the versatility of MMM projects we do. We obviously excel at the OG MMM (FMCG/CPG brands) and have got the world's best brands as our client 😎. But apart from this, we also push the envelope to implement MMM in domains that had not adopted it before. We recently implemented MMM for a major football club ⚽. 🎯 Objective: To know what are the key drivers of their crowd attendance and to study the impact of marketing /media spends on driving crowd attendance. 🛠 Approach: - We built a MMM model that answered the question of what are the key drivers of crowd attendance. - We computed the spend share and effect share of each marketing/media channel. - We then put our thinking hats on and converted their original problem into that of cost per acquisition (CPA) problem to do budget optimization. 📈 Results: ✅ The Football club clearly got to know which media channels is driving crowd attendance the most. ✅ Realized where they were over spending and under spending. ✅ We showcased the scenario where even with 15% increase in marketing budget (their budget increase for next season), we could reduce their Total CPA by nearly 5% !! Stay tuned for the detailed case study! If you're from a non-OG MMM domain, remember MMM can work wonders across sectors. And about our animated budget optimization chart, the client absolutely loved it ❤️! 

Attribution ≠ Causation In the last few weeks, I have been seeing posts on Linkedin from marketers where they describe attribution as "the action of regarding something as being caused by a person or thing." They then go on to vilify Multi touch attribution (MTA) as being a misnomer as it is not causal in nature. IMO, I think the nomenclature of Attribution is correct and people who named it, didn't mean the first result that pops up on Google. Rather they meant the second result that pops up on google when the search term attribution meaning is keyed in. The second result on google says "the action of ascribing a work". Here is how I would distinguish the two: 📌 Attribution : Ascribing or Assigning credit that is predominantly subjective* in nature and not borne out of rigorous causal experiments like Randomized control Trial (RCT) or Quasi Causal experiments like Difference in Difference (DID), Regression Discontinuity (RD) or Propensity matching. 📌 Causation : Establishing clear cause and effect as a result of causal experiments like Randomized control Trial (RCT) or Quasi Causal experiments like Difference in Difference (DID), Regression Discontinuity (RD) or Propensity matching. While ensuring we control for all the factors and take care of effects like Confounder effect, Mediation, Moderator effect and Collider effect. 🔍 What is further causing the confusion? I believe marketers are getting confused because they are drawing parallels between MTA chart and DAG of any causal exercise. DAGs are easy to draw. But just because somebody has drawn those pointed arrows, it does not mean they established causality. 😅 DAGs most often is a way to summarize the causal research findings. There is a differing school of thought whether DAGs are appropriate or not between Causal inference stalwarts like Guido Imbens and Judea Pearl. I will leave a link in comments to the intellectual battle they have had in the past - 'To DAG or NOT'. Note *: While I mentioned that attribution is predominantly subjective in nature, attribution often in case of MTA is arrived through probabilistic techniques like Markov Chain Monte Carlo (MCMC). While these are note subjective, they are at the same time not causal as well. DAG image credit from Matheus Facure's book - Causal Inference for the true and brave.

It's challenging to transition from established solutions. But holding onto third-party cookies for attribution isn't viable with Google set to introduce an opt-in system akin to Apple's ATT in 2021. Apple's ATT saw initial opt-in rates of 5-10%, and Google's rates are projected to be similar or lower! Marketing Mix Modeling (MMM) is one of the light at the end of tunnel for marketers. We are currently helping many clients worldwide cross the finishing line of adopting MMM. Get in touch with us - we can help you cross the finishing line of 'Adopting privacy proof MMM solutions' 😎

I recently stumbled upon a tweet response to "What controversial statistics take will have you like this (image of a man surrounded by men with spears ready to attack him). Ed Kroc in his responses said the following about Bayesian methods. 📌 "Bayesians should be forced to scientifically (not mathematically) justify their priors." 📌 "A prior should be scientifically/substantively defensible, otherwise why bother with Bayes?  If you're truly ambivalent about the estimand, then use a flat prior, so let the likelihood drive everything, so again: why bother with Bayes?" 📌 "Bayes is usually used as a way to "fancy up" a problem without sincerely engaging with its complexities.  I prefer the Bayesian approach theoretically, but I don't trust the average non-stat person to do it well." I totally agree with all of the above. Bayesian methods are indeed used to 'fancy up' a problem. It is used to give an impression that something deeply scientific is happening. Bayesian will showcases chart after chart of sampling from a distribution as if something ground breaking is happening 😅 . I will give credit to Bayesians for succeeding in the marketing efforts though and for creating an impression that - Bayesian methods are intuitive and easy. Coming to the field of Marketing Mix Modeling (MMM), Bayesian MMM is just used to "fancy up" the marketing measurement and attribution problem. You will hear high sounding words like 'Parameter recovery', 'time varying parameters', 'Uncertainty Measurement' etc. At the end of the day, MMM much like a multi linear regression problem is about estimation of parameters. One need not make estimation of parameters any more difficult, complex and compute intensive than it could already get. But Bayesian MMM do exactly that and it provides tremendous room for manipulation. MMMs are typically small data problem (relatively speaking). And the priors almost always overwhelm the evidence in the data. Not to mention the other big Achilles heel of Bayesian MMM - Multicollinearity. If you want a detailed explanation of the shortcomings of Bayesian MMM - the links are in comments. Lastly I find the meme highly relevant. Frequentist MMM is like Yusuf Dikec (Right). Gets the job done without any fuzz or "fancy-ing up" 😎. With Bayesian MMM the question begs - why take the complex, convoluted, compute intensive and error prone route. Link to the tweet response in comments. P.S: Just to clarify, I am using the meme in jest. No disrespect intended to the South Korean shooter.  Congratulations to both Turkish and South Korean athletes on the Silver Olympic medals.