The Right Way to Think about Marketing Mix Models

Here’s a fine mess: I was all set to review a paper titled “Marketing-Mix Modeling the Right Way”, partly because I found the title annoyingly arrogant and partly because I have doubts regarding the methodology it describes. But when I tried to find the paper on the vendor’s site, it wasn’t there. Checking my files, I found my copy came from a friend, so perhaps it had never been released publicly. That being the case, it would be improper to critique it here.

Problem is, I now have marketing mix modeling on my mind. So I’ll just have to plunge ahead even without a white paper as a hook.

What makes marketing mix models relevant to customer experience management? The basic answer is, they illustrate a way to optimize an important set of customer-related resources. This is something customer experience management must also do if it is to move past the soft notion of “be nice to your customers” to become a serious management tool.

I suspect that most people reading this blog are already familiar with marketing mix models. For those who are not, let me correct the impression that marketing mix models simply take a company’s advertising budget and correlate it with changes in sales volume. A serious marketing mix model considers competitive activities, product pricing, retail distribution and in-store promotion. (This is why it’s called a marketing mix model and not a media mix model.) An advanced model might also incorporate the contents of marketing messages and the behavior of different customer segments. All this data is evaluated separately by geographic region. Gathering the information typically requires access to external data on advertising spend, promotions and product sales from vendors like Nielsen and Information Resources, Inc.

Marketing mix models address several key concerns for customer experience management: identifying events that influence customer behavior; integrating external information; building models that capture relationships between multiple events and behavior; and identifying optimal resource allocations. Of course, marketing mix modeling only looks at a subset of all customer-related events; the operational experiences that are central concern of customer experience management are excluded. On the other hand, marketing mix models offer much greater precision than most customer experience analysis. So it’s worth asking whether the techniques of marketing mix models can be extended to support customer experience modeling, or at least can offer some useful lessons.

The problem with directly extending marketing mix models is that they are not based on individual customers. It is certainly possible to add operational metrics as inputs: say, on-time arrivals if you’re an airline or order fulfillment accuracy if you’re a distributor or customer satisfaction scores for just about anyone. This could give some measure of how experience impacts over-all results. But it wouldn’t be directly measure the impact of specific experiences on individual customers, so the information would be hard to interpret and important relationships might be hidden. I suspect different statistical methods are needed for the individual-level analysis. I also question whether tools that predict aggregate sales by period can also project customer lifetime value.

A quick scan of marketing mix modeling vendor Websites didn’t find anybody addressing these issues, although I might have missed it. (Vendors I looked at: Upper Quadrant www.upperquadrant.com; M-Factor www.m-factor.com; Marketing Management Analytics www.mma.com; Hudson River Group www.hudsonrivergroup.com; Analytic Partners www.analyticpartners.com; Pointlogic www.pointlogic.com; Marketing Analytics Inc. www.marketinganalytics.com; Copernicus Marketing Consulting www.copernicusmarketing.com; Strategic Oxygen www.strategicoxygen.com; iknowtion www.iknowtion.com; Management Science Associates, Inc. www.msa.com; ACNielsen www.acnielsen.com [recently purchased The Modeling Group]; SAS www.sas.com [recently purchased Veridiem].)

In terms of broader lessons, the marketing mix model vendors certainly offer some useful techniques for gathering and integrating external data. They typically do this by geography (related to retail trading areas and advertising markets), which is useful because most customer data can be linked to a physical location. The general skills needed to build and calibrate marketing mix models are also relevant to customer experience modeling.

Perhaps more important, marketing mix models help companies develop attitudes that are needed for customer experience management. These include understanding that models need not be perfect to be useful; using simulation to make business decisions; considering the trade-offs inherent in the concept of optimization; including external as well as internal factors in explanations of customer behavior; moving beyond a purely product-centric view of the market; and trying to measure the real value of traditionally unaccountable activities such as advertising spend.

In short, marketing mix models may or may not provide the right technical platform to build customer value models, which I consider an essential underpinning of serious customer experience management. But marketing mix models do provide a useful template for making a sophisticated analytical tool part of high-level decision-making. This alone makes them worth a look from a customer experience management perspective.