Unearth the power of declared data
Insights That Work
Brand & Retailer tickets for all IIeX events now start at just $99! Get or give one today!

The Death of Marketing-Mix Modeling, As We Know It

An attempt at outlining a vision which certainly can become the foundation for a renaissance and rebirth of this important tool for marketing measurement and marketing insight. If MMM is to survive, it is essential that it change and experience a rebirth.

Editor’s Note: Last week we posted a great think piece by Joel Rubinson on the need to focus on “bottom-up” marketing. One of the keys of achieving marketing ROI  is developing effective mix models that factor in the new world of digital attribution and programmatic advertizing. However, mix modelling has failed to keep up with advances in adtech and martech so it needs a real refresh. We believe so strongly in this that we’ve partnered with Sequent Partners, Time Inc, the MMA, and many other brands, agencies, tech platforms and media networks to develop a new event: the IIeX Attribution Accelerator forum.

Mix modeling needs to become more granular, timely and actionable – like attribution. Attribution needs to be more comprehensive, addressing the entire marketing mix, and needs more scientific rigor – like marketing mix modeling. As attribution moves beyond digital, and marketing mix modeling moves beyond traditional, a more integrated approach to marketing measurement is needed.

In today’s post, industry legend Michael Wolfe offers a tour de force on the problems with traditional mix modelling and a prescription on how to make them better in the modern era.



Despite my own hands-on involvement with MMM for nearly 30 years, I tend to believe that marketing-mix modeling, as currently practiced, is broken.  Sadly, I have gone back and looked at various vendor MMM power-points from 20-25 years ago; and frankly see little change from the current versions in terms of the issues covered and the underlying methods used.  Clearly, MMM does have a problem with being relevant and adapting to a more complex and fragmented market place.  Much of the criticisms are valid.

I, for one, think that rather than closing the coffin, this is an area in need of a renaissance and rebirth.  In order to do that, however, several things need to happen:  (1) an agreement and recognition of what the MMM shortcomings are, (2) a means or solution-path to solving these shortcomings, and (3) a concerted effort by MMM buyers to require, and MMM vendors accept, the solutions needed to achieve this renaissance.  This paper will outline what some of these solutions need to look like.

Recognizing the Shortcomings of MMM

Much of the criticisms of MMM derive from issues with the prevailing methods, the breadth of its ability to generate accurate insights or its inability to measure certain aspects of the marketing ecosystem.  When combined together, these are all saying that MMM is losing its relevance to answer critical business questions and has failed to keep up with the growing complexity and fragmentation of the market-place.  The interesting thing, however, is that I believe there are current solutions to most, if not all of these short-comings.


  • MMM focuses on the short-term effects of media and generally ignores or does not measure the long-term effects.

Interestingly, some MMM buyers use a rule of thumb 2X multiplier to adjust short-term ad incremental sales.   This is a cop-out.  Based on experience, these effects range from 1X to as high as 6X.


As shown below, the long-term effects significantly expand the total measured impact of all media. This is not a lost cause because there are methods for measuring these long-term effects.  Every MMM should directly derive this for every brand.


The significant value-added from measuring long-term media effects is that it often completely changes or reverses the economics of advertising.   In failing to address this, historical MMM has led companies to misallocate resources.  Often to the detriment of long-term equity building (advertising) for short term gains (e.g. excessive price discounting).

It is also said that most advertising for CPG companies loses money.   A study by the firm MMA, for example, finds that, over hundreds of brands, advertising only returns about 54 cents for every dollar spent.  However, as shown in this case, the long-term ad effect transforms the ROI of media from negative to strongly positive!  When the results of any MMM show negative returns from an activity that is actually positive, this can result in a misallocation of marketing resources going forward and sub-optimal business results.


  • MMM models only measure the impact of ad GRPs or spend and not the ad message or creative.

Whereas media channels such as TV or Digital are the delivery vehicles, the ad creative content or message is really the life-blood of advertising and how brands connect with customers.  MMM generally  fails to help advertisers understand the relative importance of message or creative content in driving brand performance.


Yet, it is quite possible, with some ad copy-test protocols, to scale the ad spend or model GRP variable with the copy test scores and, thereby, permit the separation of advertising “creative effects”  from “weight or spend effects”.  Below is an example of this.

While the mechanics and analytics behind this are all quite feasible, the challenge in doing this is that it requires the copy testing of all ads.   Copy testing is expensive and it is also not certain that every copy test method or regime will work effectively in this context.  Nevertheless, there is great value and insight gained from understanding and monetizing the value of ad creative or message. Despite all of this, there are affordable solutions. Measuring and monetizing the value of ad creative changes the focus of marketing more towards the message, where it should be.


  • MMM does not account for attribution bias, particularly within digital media.

MMM becomes confused when customers are affected by a pathway of marketing touch points, rather than a single exposure.    Therefore, MMM will be biased with respect to giving significantly more credit to that marketing channel in which the customer touches just prior to purchase.


Despite this, there are “path modeling” methods that treat marketing touch points as separate equations.   This shows great promise for overcoming this obstacle.  It also makes a major difference in how different digital media are estimated to affect purchase, as shown from the chart below.


  • MMM tends to not quantify the “synergies” across media channels, where the impact of two or more simultaneous media activations is greater than the sum of the independent parts.

The underlying principle behind econometric models of MMM is that variables or drivers are assumed to be “independent”.   Unless special interaction terms are built in or other non-regression based methods are used, all media in a model are assumed to be unrelated to the others.  We know this is not true.


As shown below, it is often the case that when two or more media are executed simultaneously, a greater impact is found.  The total is here greater than the sum of the independent parts.  As shown on the chart below, we call this “marketing synergy”.   Currently, we have found a lot of these synergies between digital and more traditional mass media.  Some of the current criticisms of MMM are that it underestimates the total impact of digital media and this phenomenon is a major factor explaining this underestimation.  By measuring these synergies, marketers will see the marketing mix as a symphony of instruments that need to work together harmoniously to maximize revenue.   This is the basis and measure of the value for “integrated marketing”.


  • MMM modeling might be able to explain what is happening to brand sales, but because it excludes the “voice-of-the-customer”, it fails to explain “why” and provide insights based on the customer’s mindset and current brand experience.

Failing to understand the mind and voice of the customer, in context of a MMM exercise, is a serious and nearly fatal weakness.   That is because, without understanding the voice-of-the-customer, it  is really next to impossible to understand the “why’s” behind a brand’s performance in the market place.  This also fails to consider how any decisions or changes to the marketing mix will be received by the customer or consumer.


One very effective way to address this issue is to impute a highly predictive social media metric into the MMM.  Below shows the sales correlations of one such metric.  In this particular instance and across 28 brands, the correlation of the Social Engagement Index or SEI is quite large.  This gives it plenty of leg-room to be included in a predictive marketing model.


By including the voice-of-the-customer into MMM, one gains a depth of insight that enables a deeper understanding of what part of the customer experience is most affecting brand performance.  As shown, in the case of a restaurant brand, issues with their menu and perceived  food quality were driving much of this company’s sales decline.  From the customers own words, this provides an explanation of the “why” behind brand performance.

Probably the most important reason for including the VOC into a MMM is that, as it represents the customer-brand-experience, its overall effect on brand sales is usually one of the largest business drivers.  It is even sometimes greater than that of all other marketing factors combined.

Reinventing Marketing-Mix Modeling

While there has been a lot of change in the complexity and elements of the marketing ecosystem over the past 30 years, the prevalent marketing measurement tool, Marketing-Mix Modeling, has failed to keep up and adapt to these changes.  In short, for the reasons cited, MMM is broken.  We are now at a cross-road.  Either MMM can continue to slowly die or else it can truly adapt and evolve.  As shown here, the solutions for overcoming its short comings already exist.  To achieve that renaissance, MMM must:

  • Expand its measurement focus towards quantifying the longer-term effects of marketing and develop more accurate and holistic ROI estimates.
  • Focus more on measuring the effectiveness of ad messaging and creative to better align and develop marketing communications strategies.
  • Adapt its method to avoid the pitfall of “last touch attribution”
  • Measure the interactions and synergies that exist between and across the marketing-mix, in order to form a foundation for integrated marketing.
  • Put the “voice-of-the-customer” front and center within the models in order to more fully understand the customer’s perspective & motivations driving business performance.

It seems that much of these changes have not been forthcoming from the MMM vendors.  Therefore, the key to effective change rests with the advertisers and buyers of MMM to require that these changes occur.  Without changes of the nature outlined here, there is a risk of MMM becoming obsolete and commoditized to the point of having little value.

This paper has been an attempt at outlining a vision which certainly can become the foundation for a renaissance and rebirth of this important tool for marketing measurement and marketing insight.  If MMM is to survive, it is essential that it change and experience a rebirth.

Please share...

3 responses to “The Death of Marketing-Mix Modeling, As We Know It

  1. Very interesting article. I did wonder about one comment – the predictive power of sentiment analysis and sales. When you look at the charts I cannot see that. The two lines seem to be totally correlated raising the issue of whether it is purchase that drives sentiment not the other way around. To be a good predictor one would expect some kind of lagged effect.

  2. I think we need to start again and model from the bottom up with Agent Based Models

    Agent based modelling as an alternative to marketing mix modelling is decribed here

    “The key distinction is that marketing mix analyzes a single brand’s activity in aggregate, while agent-based simulation recreates a brand’s marketplace based on individual consumer actions.”

    “So in summary, marketing mix modeling is an approach for performing optimization of paid marketing investments for one brand in a stable environment. Simulation explores a broader range of strategic options that consider the competitive market and consumer behavior. So the question on how marketing mix compares to simulation always comes back to what questions you need to answer as a decision-maker.”

Join the conversation