I'll be honest. I've gone back and forth over the years on the value of marketing mix models, all the time seeking to understand where, when, and how they add the most value to marketing managers looking for insight into more effective and efficient marketing.
I think I've seen all the warts:
· Models that just skim the surface of tactical optimization by looking only at media and ignoring the category dynamics and uncontrollable factors, thereby overstating the impact of marketing in very suspicious ways.
· Models that look backwards perfectly but are of little value if one is making decisions about what to do next.
· Models that regularly seem to reward short-term demand-generation tactics because they can "read" those more cleanly than longer-term brand equity value.
· Models with high "base" compositions that explain only a small portion of the change in volume or profit from one period to the next.
· Models that understate the value of marketing by failing to account for the interaction effects, particularly those in the new digital and social media realm.
I've also heard some sophisticated marketers who have long used these tools question their predictive value, particularly in complex competitive ecosystems with lots of channel influences and/or short innovation cycles. I've heard many voices I respect in both business and academia say that MMM is "mature" (as in "aged," "long in the tooth," or "declining in value or relevance").<
So you'd think that I'd have given up by now and moved on to something more innovative that might address some of these issues like artificial intelligence; agent-based modeling; systems dynamics, etc. Nope. Call me a MMMM (Marketing Mix Model Masochist) if you will, but all this criticism has actually led me down the path of embracing them more than ever. Principally for one simple reason: they are effective ways of helping managers understanding what has happened and what might happen looking forward.
For whatever reason, MMM is a concept that most marketing and finance managers can actually grasp. It is neither too simplistic to adequately explain their understanding of the universe they operate in, nor too complex to be embraced and acted upon. MMM is sort of the Goldilocks solution: just right.
Sure, there may be other techniques that could answer any specific question more comprehensively, but at what cost to transparency and credibility? Managers (at least the human ones) need to 1), understand, and 2), believe the analysis they get in order to operationalize it. The core mathematical simplicity of MMM gives it an advantage in both those respects.
But I'm pretty sure that MMM is "mature" only in the sense that it is a settled science which keeps getting better with age. In fact, using lifecycle terms, I'd categorize MMM as more in an "adolescent" stage. New methods and techniques have emerged to help better decompose the base effects, isolate long-term brand impacts, explain the direct and indirect effects of various elements of the tactical spectrum, and improve the forward-looking capability of the models.
When combined with some appropriate adoption training and alignment within the management team, today's marketing mix models actually:
· Provide more operational guidance, aligning increases or decreases in marketing spending with channel management and supply chain considerations;
· Link to trade-off analyses on a market segment or brand equity level; and
· Help establish spending strategies as market conditions change.
Improved automated functionality is also allowing marketers to react more quickly to results based on their needs -- by refreshing the models frequently and putting broad "what if" simulation at the hands of the marketing planner.
Ever the skeptic though, I still see some need for improvement. Specifically, MMM cannot be effective unless the modelers:
· Ensure that the organization as a whole understands the assumptions and limitations of the marketing mix model;
· Realize that laying the acceptance groundwork around those assumptions is as important and challenging as building the algorithms or collecting the data;
· Remain aware of changes in the competitive environment and how they affect business results; and
· Understand that the model will, inevitably, fail; expect this event and plan for it.
Finally, I think it's critically important that marketers NOT stop at marketing mix models. Risk is magnified by over-reliance on a single tool. Today's marketing measurement toolkit needs to be much broader. Deep understanding of brand drivers, customer behavior and value require input from tools and techniques outside the mix model, as well as in.