Marketing Mix Modeling (MMM) – Optimize your Marketing ROI

Marketing Mix Modeling is greater than Media Mix Modeling

Marketing Mix Modeling > Media Mix Modeling

Marketing Mix Modeling > Media Mix Modeling In a recent client pitch meeting, a client asked us what is the difference between Media Mix and Marketing Mix Modeling. Quite clearly, the client was confused because some vendors market their services as Media Mix Modeling instead of Marketing Mix Modeling. So, is this just a word play or is there more nuance to it? In my opinion, there is more to it than just word play. While, Media Mix Modeling is focused only on understanding advertising impact on the brand’s sales/market share/leads etc., marketing mix modeling involves understanding other aspects as well like impact of price changes, distribution, promo offers and

Learnings from OG MMM – FMCG/CPG MMM

Learnings from OG MMM – FMCG/CPG MMM It is a well known fact that FMCG/CPG industry were the pioneers in adopting Marketing Mix Modeling (MMM). My initial training was in applying MMM to FMCG/CPG marketing effectiveness problems. In the last few years, MMM has had a renaissance of sorts (courtesy various factors like privacy, cookie deprecation, GDPR etc.) It is now being readily adopted by many different domains. Here is how the learnings from FMCG/CPG MMM helped me to apply MMM to different domains. 1. Observing the effect for longer time period In FMCG/CPG MMM, one typically sees data of past 3-5 yrs. This helps one understand how the various media

Weekly MMMs?

Weekly MMMs? Some prospective clients ask us: ● How frequently should they update their MMM models? ● Is a weekly cadence ok? Well it depends. There are business reasons as well as technical. 📈 Business reason: Whether weekly MMM is ideal depends on the industry. If you are a digital focused brand with high frequency of short term campaigns, then it makes sense to update the models weekly or monthly. But if you are focused on brand marketing, you would not be able to see substantial changes in weekly updates. Most brand building activities take time to reflect in the incremental sales. 🛠 Technical reason: Coming to the technical reasons,

How MMM can make the CMO role a ‘Non Revolving Door’ job

How MMM can make the CMO role a ‘Non Revolving Door’ job. ● “Can you help us prove that our marketing efforts worked?” ● “Can you help us find out which marketing channels are working, and which are a drain on our budget”? ● “How can we increase or maintain the same Marketing ROI (MROI) if the budget is constant for Year 2023”? These are the typical questions I often get from CMO’s. CMO’s often get blamed for the downside (sales going down, leads not trickling in, not enough CTR etc.). But rarely a CMO is given a pat on the back for the upside (Incremental sales, more leads, higher

DIY MMM ?

DIY MMM ? Of late we are getting lot of calls from prospective clients for MMM adoption. One of the curious question being “Can we DIY MMM through open source libraries?” Our answer – Yes and No 🙂. You can DIY MMM if: ✅ You understand the nuances of MMM (Remember MMM is not just Linear Regression). ✅ You understand the domain. ✅ You have good knowledge of statistics. ✅ You have good understanding of experimentation and causal inference. ✅ You know how the open source libraries work under the hood. ✅ You know R or Python to make changes in the open source libraries if required. If you don’t

Saas-ification will not fix the core problems in MMM

I increasingly see that more is being talked about ‘How MMM can be done in 1 day’ rather than how the problems of multicollinearity and endogeneity are solved. Saas-ification in parts is good and even desirable. We ourselves have a SaaS platform ‘ArymaEdge’. But our SaaS platform does not try to fully automate the MMM process. We still build the MMM the old fashioned way because the problems of multicollinearity and endogeneity can only be addressed through careful econometric modeling. Why is the problem of multicollinearity and endogeneity so crucial to address? Lets take an analogy. Marketing attribution is a problem of identifying the Most valuable player/s (MVPs) in a

Short term sales activation vs brand building measures?

A lot of D2C brands often want to benchmark their marketing effectiveness against other D2C players, disregarding the product category. While a similar marketing effectiveness strategy could work for fast moving goods, the same cannot be applied to D2C brands with longer sales cycles – like furniture or electrical appliances. I have seen some brands taking cue from D2C FMCG products and think that the performance marketing measures working for these fast moving products, could also work for them. In some cases, they also want agencies to prove a high short term ROI or rather prove the efficacy of the whole Marketing effectiveness exercise by citing examples of these brands.

Quantum of Marketing Spend vs Pattern of Marketing Spend

In every Marketing Mix Modeling (MMM) project that we undertake, we always study the previous spend pattern of the client. Why? Because how much one spends in marketing channels gives us only partial information. The other half is about the spend pattern in those channels. A brand with a ‘single burst’ strategy may not be able to match the ROI performance of a brand that has been investing in channels consistently. A haphazard spend pattern also affects the Reach and Frequency. Therefore we always conduct a Spend pattern analysis before building an MMM model. Below are the benefits of spend pattern analysis: 📌 Gives us a clue about the client’s

Do you use the right tools to evaluate your MMM model?

Do you use the right tools to evaluate your MMM model? Having the right tools to measure your MMM model’s accuracy is as important as specifying the model correctly. Contrary to popular belief, MMM is just not about retrodiction. If the model is well specified, it can also be used for forecasting (for a reasonable future period). This comes handy in media planning and marketing budget planning. Forecasting is also about picking the right accuracy metric. Here are some pointers that may help you to choose between MAPE and RMSE. MAPE – They cannot be used for intermittent patterns where the observed value (e.g. demand) becomes 0. RMSE – RMSE

Using Transfer Entropy for Feature Selection in MMM

Few months back we published our whitepaper ‘Granger Causality – A possible Feature Selection Method in MMM’. ICYMI the link is in the resources section. In this blog I want to highlight another useful feature selection method – Transfer Entropy. As you must have guessed, the method is a information theoretic method. In case you would like a good refresher on topics of entropy and information gain, I would highly recommend the article ‘Entropy Demystified by Naoki’ (link in resources). So coming back to Transfer Entropy. Transfer Entropy is a non parametric method to measure the amount of directed (time-asymmetric) transfer of information between two random process. Transfer entropy from

Blogs