Commentary

Marketing-Mix Modeling Doesn't Go Far Enough To Solve The Attribution Riddle

Google’s move last week on marketing-mix modeling represents a small step in supporting independent measurement. Google said its Marketing Mix Model (MMM) Partners program will offer reporting and service for a few companies including Ipsos, Neustar’s Marketshare, Nielsen and other vendors who apply.

“I was pleased to see Google announce support for independent measurement platforms. Knowing that Google has a multi-touch attribution (MTA) product of its own which aspires to be the record of measurement, is an acknowledgement that MTAs, including Google’s, have failed to provide reliable attribution beyond digital," Rex Briggs, founder and CEO of Marketing Evolution, told RTBlog via email. However, Briggs said he hopes marketers will go beyond MMM to embrace “cross-channel, person-level measurement, which offers much more detailed measurement and actionable optimization, and allows marketers to connect their branding and direct response measurement holistically."

advertisement

advertisement

Google itself conceded that marketers continue to use MMM due, in part, to the current limits of MTAs, which measure digital advertising, but don’t include other media channels. Briggs thinks most attribution models aren’t detailed enough or accurate. He cited a data point from a January 2017 report from DemandWave, "2017 State of B2B Digital Marketing” that 75% of marketers either don’t use any attribution at all or only use multitouch-based attribution.  

Further, Briggs maintained that most attribution models lag on indicators such as sales. When marketers use so-called “people-based” in-flight optimization, the average ROI increase is 31%, according to Marketing Evolution’s internal data. The message here is that if marketers want to optimize quickly, they need to use a leading indicator model. Most models don’t offer results for months, so marketers can’t optimize while campaigns are still running. Clearly, this isn’t helpful.

While Briggs finds MMM somewhat useful, it has big limitations, including the inability to show timely campaign results, and an overemphasis on behavioral-centered data. For example, brand equity is excluded from MMM. This can lead to overspending on lower funnel activity like search or low-cost display advertising, and under-spending for brand development among people who need more education and motivation to act.

MMM tends to ignore the role message and targeting play. If you think of media as a container holding a message that’s delivered to a specific audience, what happens when you change the message, or change the people the message goes to? Briggs maintains that mix models miss how advertising works and thus, fail to optimize the message and targeting first, before recommending the best media mix.

Finally, MMM models aren’t detailed enough. For example, MMM puts impressions into groups like Facebook or prime-time TV, which can result in a lost optimization opportunity.

Google’s support for MMM is unlikely to address this gap. “To its credit, Google also provides fairly decent support for independent person-level measurement, at least for desktop. We have been encouraging Google -- as well as other publishers -- to improve support for mobile app video, which is currently a hole in Doubleclick’s reporting.  They’ve indicated that they are making progress on this but timing is still TBD,” Briggs said.

Next story loading loading..