The following is a guest blog from Jim Griffin, Americas Director for Cartesian Consulting and co-host of the upcoming webinar The MMM Playbook: Leading and Managing a Marketing Mix Modeling Project (Thursday, July 20).
The store-sales measurement tool that was unveiled by Google on May 23, is the latest salvo in the Google-vs-Facebook attribution arms race, and it is a big deal. In fact, this may turn out to be the most important announcement of the year from Google.
70% of all card transactions
At Marketing Next 2017, Google said that it now has access to data for about 70% of all U.S. credit and debit card transactions. Google also announced that it is rolling out a marketing mix modeling partners program based, in part, on that data. Taken together, these developments suggest that the approach and methods for marketing mix modeling (MMM) could soon be much simplified, resulting in wider adoption by advertising executives.
This latest initiative from Google has been described by some as a competitive response to Facebook’s own announcement of its Offline Conversions API which leverages similar partnerships with POS systems like Square and Marketo to pull in real-time results of offline transactions. That initiative, in turn, followed in the wake of a Google AdWords upgrade to its Estimated Total Conversions store visit measurement capability. . .
What this is all about
Marketing Mix Modeling helps to identify which portion of sales is due to advertising and promotion, and which portion is due to a baseline, driven by longer-term factors such as brand equity. A related concept is attribution modeling, which is a family of techniques that seek to give appropriate credit to any of the digital channels that might have contributed to an online sale. The data context for attribution-modeling tools like Google Analytics has usually been channels like Facebook, AdWords and web referrals. By comparison, the term marketing mix modeling is broader, embracing not only online channels, but also offline channels like print and television. Using these definitions, we can say that attribution modeling has already been widely adopted by marketers in recent years, but true MMM is still an aspirational goal for many brands.
“Half of my money is wasted.”
It was around the turn of the last century that department store and advertising pioneer John Wanamaker famously observed: “Half the money I spend on advertising is wasted. The trouble is I don’t know which half.” – a turn of phrase that is still often quoted to this very day. Somehow, the elusive promise of MMM was to finally identify whichhalf was actually wasted, and to fix that.
Almost 30 years ago . . .
The first documented commercial efforts towards MMM took place back in 1989, about the same time that Tone Loc was singing about the benefits of Funky Cold Madina. That’s the year that Hudson River Group was founded, followed in 1990 by MMA, which was then bought by Carat, then acquired by Aegis, then rolled up into the Synovate division, and then spun off. Somewhere along the line, MMA founder and industry icon Ed Dittus disappeared from everyone’s radar, pivoted, and reemerged as a designer of rocket launchers and guns.
Other early pioneers had their own stories of reincarnation. ATG Group was acquired by J Walter Thompson, and then became part of Mindshare. BrandScience became part of Omnicom. Skunkworks was acquired by Bozell Jacobs, and so forth. In the early years, work was mostly directed at CPG companies, facilitated by the accessibility of syndicated sales data from Nielsen and IRI, and fueled by the large advertising budgets of companies like P&G, Kraft, Coca-Cola and Pepsi.
But no one said this would be easy
Despite the insightful advice of writers like Avinash Kaushik, users of online attribution models still frequently default to flawed methods like last-click attribution. In addition, repeated waves of announcements about inaccurate data have sometimes created an atmosphere of mistrust and questions about transparency, made even worse by the existence of click fraud, especially for desktop video ads.
Meanwhile, questions also emerged about marketing mix models themselves, which some have described as a black box. To some extent, this point of view is understandable. In many cases, the key players in MMM were treating their methods as intellectual property, not to be revealed in too much detail. Each one was essentially claiming that only they had the right approach – and that it was a secret.
And along with the lack of clarity about best practices, questions have also been raised about the appropriate objectives for MMM. In 2009, for example, Bill Harvey of TRA Global cited the experience of Mars candy as evidence that differences in creatives might actually be more important than all other factors combined, including marketing mix. In reality, this is probably true. The proponents of MMM never claimed that creative was unimportant. The question they were asking was different: Once the creative is selected, how should spending for that be allocated across channels? Nonetheless, the pesky questions remained. How, for example, should MMM methods deal with a dataset that includes not only different channels, but also different creatives?
S*** just got real
As recently as 2011, smart people like Laura Desmond of Starcom Mediavest found it reasonable to predict that market mix modeling “will become obsolete,” although Laura admitted to being intentionally provocative in saying so. In my opinion, the latest announcement from Google takes this prediction off the table as a plausible scenario. The ability to match online behavior to POS transactions will surely bring in a whole new wave of marketers that may have looked at MMM as too difficult or too esoteric before.
Perhaps data was never the real problem, nor any of the other many criticisms that have been leveled at MMM over the years. More likely, the real stumbling block has been inertia all along, augmented by unfamiliarity with the art and science of MMM. In many cases, the real question was and is: How does a non-technical marketing leader successfully orchestrate a highly-technical MMM project? That, in fact, is actually the topic of an interesting upcoming webinar, to be co-hosted by my colleague Tapan Khopkar, PhD. (Tapan leads the Advanced Analytics practice at Cartesian Consulting, including MMM.)
One way or another, marketing mix modeling seems positioned for another Renaissance. The innovators and early adopters have already blazed the trail, and the latest announcements from Google and Facebook will likely usher in an early majority phase. Needless to say, when it comes to marketing, late majority is not a good place to be. In short, if you’ve been thinking that you should “look into MMM someday,” then maybe someday has just arrived.