eclecticallysampling.com

Eclectic Thought Samples From Larry Burns

Is Our Market Moving Too Fast for “Classic” Marketing Mix Models?

Share Button

As the utility of data from consumer interactions with brands, retailers, online shopping, social media tracking, email tracking, etc. continues to expand, the concept of our industry’s marketing mix models has become an ongoing concern for me.  I am talking about macro-level models that are still used to allocate budget monies in marketing budgets.  (To be clear, I am NOT talking about myriad models running in real time for things like microsecond ad auctions for ad display, loyalty program optimizations, personalization routines used to create what is loaded to a web page you arrive at, or the literally thousands of mathematical models deployed to interact with online tracking tools.)

“Classic” Marketing Mix models were initially created back in the ‘70’s.  The noble intent was to try to help marketing professionals intelligently determine and budget for what percentage of their resources ought to be spent across the multiple, yet limited forms of media that existed.  Then, in the late 80’s, in an early burst of data explosion (when 10 terabytes actually was HUGE!),  the integration of actual sales data within the models began.

What marketing mix models do, in layman’s terms, is a very sophisticated  “look back” to examine what influenced sales trends, layering in all manner of marketing variables (media weights, price, promotions, etc.), with a desire to “tease out” the effects of each of these marketing elements.  Algorithms are created to reflect the ‘‘weight’ or impact of various forms of media, promotion, etc. for that particular brand and category.  Then, remarkably accurate, sophisticated mathematical models are created to offer guidance for future spending.  “What will happen to my sales next quarter if I move $5 million from my advertising budget to my shopper marketing budget?” is a tempting and very legitimate question for a senior executive to ask – expecting an answer.  He or she will get one from their team, and on they will go… I simply say that, collectively, we need to take a good look “under the hood” at these models, even if we feel ignorant in the face of data gurus who, by spouting complexity, can silence questions because many otherwise very talented people simply do not know what to ask.

I was around when the first generation of scanner sales based models were created.  In Westport, Connecticut, in my role at Cadbury Beverages, I fed IRI data to MMA around the first time such data was effectively integrated.  Now, anybody who is honest in creating these models will admit to you (in particular, over drinks in the bar AFTER the presentation) that there’s a definitive ART to this as well as the science of deep, complex mathematics.  The science has advanced rapidly with faster, more powerful machines that allow us to sort through vast quantities of data (about to pass a Zettabyte or 1021) now being created.  Many retailers have entered into agreements with major information partners to examine their storehouses of data relating to actual transactions at a person/store level, in order to explore how behavior can be “predicted” based on history. (See Machines of Marketing for a rant on that)

While I am no longer using, buying or selling these models, I struggle to believe that traditional marketing mix models have been updated with sufficient speed or precision to legitimately include all of the new media and myriad ways in which consumers (or people as I like to call us) are interacting with brands and retailers today.    No one is at fault – the pace of change is simply too great.

However, it is still far too common to engage with literate, intelligent and innovative marketers who, when presented with alternate, potentially disruptive strategies to engage people in the marketplace, respond with something that begins with:  “Well, our model says…“  This can be very frustrating for those of us who are seeing fascinating results as we create new strategies and tools to address the changing shopper as she moves along at a pace of change greater than one to which even terrific marketing staffs can adapt.

The marketplace has always been in a constant state of evolution, and I will avoid boring you with examples, as most of you know about marketing changes over the years.  I also know and greatly respect that in the marketing sciences community, there’s great work underway to integrate massive change in ways that help decision-making.  But at the end of the day, I ask a simple question.  “Has our corporate reliance on trying to minimize the risk of a wrong choice actually increased the opportunity for our competitors?”   

From my seat for this past decade of “marketing practitioner,” it is abundantly clear that the old paradigms don’t work, and we have not, as an industry, adjusted rapidly enough.  Recalling the steep uphill battle my current company faced from 1999 until nearly 2005 with adoption of the web, I also am acutely aware of how much faster adaptation has been taking place in this century’s second decade.   My bias is to believe in marketing mix modeling as an input to resource allocation choices – I’m an old “Marketing Research guy” (with the increasingly white beard to prove it) who does truly understand the value of these tools.  Yet there lives an underlying unease in me, as but one steward of my customers brands, this lingering feeling that too much reliance on “our models tell us” can be used as an “easy answer…” to avoid the inherent risk of “the new.”

The way that people shop and interact with each other, let alone with brands, is changing far more quickly than my CPG industry can adapt.   Good people across our industry are trying to adapt, and many are doing so with startling rapidity compared with, let’s say, the 80’s.  There are abundant, very clever vendors offering all kinds of interesting solutions to help the process move along faster.

It just remains challenging for large CPG companies to leave behind successful, profitable strategies they’ve used that are in some cases sustained by the inertia of money flow.   Many of us know certain marketing devices will not be with us much longer and yet they persist when innovative thinkers have been pronouncing them dead.  Why do they survive?  Guess what – the models still say this tactic “works.”  In our world, where weekly scrutiny of sales trends is coupled with the plague of daily stock price (a poor gauge of anything fundamental to the business), these pressures conspire to diminish management’s ability to take a longer, and possibly more risky view to enable sustainable success.

Looping back to my opening, let’s hope that we’re asking the right questions of our talented modelers, and demanding ongoing proof of the continuing value of the marketing mix models we deploy.    Given the oft forgotten reality of the Art embedded in the assumptions that every model contains (in varying degrees), we owe it to ourselves and the industry to challenge such assumptions.  So, ask we must in hopes an honest discourse can occur.  By questioning model assumptions, we are not attacking nor attempting to discredit the model and avoid the reality of the math.  Rather, with greater clarity on how the answers that we are going to use are derived, we can then apply judgement along with model results to reach the best decisions.

Today, when marketers can hold ourselves accountable via ever-increasing data flow, we should more fully comprehend the scorecards being used.  My final question: “How do we present this so that it makes sense in the 15 minutes that we might get to spend with a CMO or even a CCO as some think they ought to become?”


Categorised as: Learning, Marketing, Trending News


  • <blockquote data-in-reply-to=”291723514917048320″><p>@<a href=”https://twitter.com/jlarryburns”>jlarryburns</a> Nice blog, Larry. As in the 70s (I’ll trust u on that one) we model only what we can measure, not necessarily what is important</p>— John LaRocca (@jplarocca) <a href=”https://twitter.com/jplarocca/status/291950656921694209″ data-datetime=”2013-01-17T16:50:35+00:00″>January 17, 2013</a></blockquote><script async src=”//platform.twitter.com/widgets.js” charset=”utf-8″></script>
     
    Great comment from John “we model only what we can measure, not necessarily what is important”   I could not agree more …

  • Great comment from @jplarocca   “We model only what we can measure, not necessarily what is important”   I agree and believe that this “nuance” is oft times over looked.      Thanks John!
     
    Tweet Below:
    @JLarryBurns Nice blog, Larry. As in the 70s (I’ll trust u on that one) we model only what we can measure, not necessarily what is important