• by Ray Schultz , Columnist, February 21, 2017

An anticipated white paper on attribution and return on investment was released last week by the Coalition for Innovative Media Measure (CIMM) and the 4A’s. But it may disappoint marketers who expect it to answer all their questions about measurement. The paper concludes with “current best practices,” but admits that suggesting a single solution would thwart development of this growing discipline.

Then why bother with a white paper? Because “there’s a lot of confusion in the industry about the whole area,” said Alice Sylvester, a co-partner in Sequent Partners, which authored the study, in an interview.  

One problem is that there are two distinct forms of measurement. The first is marketing mix modeling (or media mix modeling), an older practice described in the paper as “the application of regression and other statistical approaches to estimate the impact of various marketing elements on incremental sales.”

Then there’s attribution, “the process of assigning credit to the marketing stimuli consumers encounter along the path to ‘conversion’ — taking action, sales etc.” This is “sometimes driven by ‘rules’ or algorithms, sometimes by statistical models,” the paper states.  

These two disciplines are “coming closer together,” the paper continues. “Mix models can now incorporate attribution and attribution is striving to incorporate media and marketing events beyond digital.”

The trouble is that attribution refers only to the “full spectrum of online touchpoints,” not to offline media like TV and print, Sylvester added.

Cross-channel attribution is “very self-contained” because it covers digital channels only,” she continued. “It gets more complicated once you leave the digital eco-sphere.”

Why? Because current measurement schemes focus on media, not on creative and other elements that contribute to sales — “pricing, promotion, direct marketing, PR and everything managed by the marketer,” said Jim Spaeth, a co-partner in Sequent Partners, and co-author.  

Nor does it consider externals like weather, national events and how advertising works — for example, its lingering effect. In general, “the digital people never incorporated any of this stuff,” said Jane Clarke, CEO and managing director of CIMM.

Still, cross-channel attribution is growing in popularity. As the paper states, a survey of marketers by the Mobile Marketing Association found that 39% are using it, and 67% expect to in the next three years, Sylvester said.

But what if marketers could understand “fully streamlined individual touchpoints” across all media, and make mid-course corrections as needed in real time? That would be  “nirvana — the epic promise,” Sylvester said.

At that point, they would operate at “the granularity and tempo of decision making,” she added.  

That said, the measurement challenge has to be placed in historical context. Whatever its strengths, media mix modeling was far from perfect.

“If you go back two decades ago, there was one statistical technique used,” said Spaeth. “There might have variations but it was like a wrench, an adjustable wrench, a crescent wrench, but still a wrench.”

Yet there was one group that could measure television or any other medium, only it didn’t get credit for it in the general ad world. “We would look down our noses at direct marketers,” said Spaeth.

That has changed. And the process is faster than it once was. “The cycles used to be measured once a year, like a report card,” Spaeth continued. It can now be done much more frequently.

For this white paper, the authors found 27 techniques in use — “some great, maybe some terrible,” Spaeth. They also talked to numerous agency people, and studied RFIs from 21 solutions providers.

Any advice? The paper recommends asking “exactly how the macro-level mix model and the micro-level attribution model are linked,” and how they know it works. 

These practices are “evolving, especially in digital,” Clarke concluded. Instead of reaching for perfection, marketers should focus on the current practices listed in the white paper:

• Operate at a level of granularity that enables marketing plans to be optimized mid-course.
• Use a statistical model to infer causality for all the elements of the marketing plan.
• Incorporate a baseline and all the drivers of consumer purchase behavior that account for most of the variance in sales or other outcome variables.
• The data used for all variables must be representative of the brand’s business, especially after being integrated through matching or imputation.