Attribution Accelerator Day 3: Improving Marketing Model Approaches

  • by , Op-Ed Contributor, November 18, 2021

Day Three of the Advertising Research Foundation’s (ARF) “Attribution & Analytics Accelerator” conference Wednesday provided forthright answers to a fundamental question posed by Association of National Advertisers’s Nathalie Bordes: “With the increasing difficulties and sophistication in an already complex arena, what are the keys to improvements and enhancements in the various marketing model approaches?” 

Respecting the privacy of consumers “always” was underlined by all of the conference’s presenters, including Nestle Purina’s Emily Weishaupt, who said marketers need to go a step further and embrace an explicit value exchange with consumers who share their data, and consequently, provide the first-party “gold” for marketers to model with.

Weishaupt, and her “data scientist,” DISQO’s  Stephen Jepson, cited their use opt-in panels to understand behavior across digital channels with a subsequent advanced test versus control approach.



Weishaupt clearly championed transparency in dealing with consumers in a way that builds respect and earns consent for any information exchange. And I am sure that all Purina’s end-users -- its customers’ pets -- are also particularly appreciative.

“Taxonomy,” opined, Neustar’s The Knowledge Lab’s Marc Vermut, adding consistency of insights delivery with a company’s business practices is also important.  His earlier presentation focused on the importance of people, process, culture, and data analytics to help build a unified marketing measurement ecosystem for future planning.  

Building on taxonomy, Mass Analytics’ Ramla Jarrar stressed protecting data quality via standards and ensuring data comparability is paramount. Presumably, this includes terminology, definitions, and derivation of media metrics which are fundamental inputs to any marketing model?  

She said Mass has developed a process dubbed “contemporary marketing mix modeling” (CMM), which has become a popular alternative to multi-touch attribution (MTA) models in a “post-cookie” world based on its focus on customer segments and the depth of insights it can provide.  

The array of different media “pressure” inputs used by modelers over the three days of this conference has been puzzling to me, and it would appear to be one of the improvements the ANA’s Bordes is seeking to solve.

Based on the recent research from the U.K. and Australia, anything less than an “attention” measure would surely make media findings from any model specious?  According to such research, no attention, no outcomes.  Ad spending, content rendered counts, circulation/traffic, log files, etc., are not measures of attention or even good surrogates.  

It was therefore no surprise that Cadillac’s Eric Neville indicated the automaker is moving from ad spending inputs to media audience data by channel inputs in its modified MMM approach.  However, this enhancement begs a further question.  How are the various cross-channel audience metrics defined and are they truly comparable?  

Raised during Tuesday’s session, the relevance of “data clean rooms” were cited by Marketing Evolution’s Michael Cohen as an important way of extracting privacy-complaint data from people.  Earlier he focused on media ROI (return on investment) analyses, which many practitioners do.  

However, NCSolutions’ Leslie Woods’ consistent findings on the influence of creative versus media is surely a vital consideration?  Or is it an anomaly due to typically unsuitable media metric inputs?  

Accessibility of granular data, data ingestion, accumulation and fusion of disparate data sets will unquestionably bring improvements to modeling, and the consequent findings according to In4mation Insights’ Keith Wulinsky, who presented with Dunkin’.  Speed of generating results, also a key, will improve with automation of the entire data and modeling process.  

Like Cadillac, Dunkin’ is also moving to more granular media data inputs to optimize media mix investments and address its local and regional ad delivery concerns.  Interestingly, the marketer uses both MMM, as well as attribution approaches.  

Jim Spaeth, partner at Sequent Partners, which co-hosted the conference with the ARF, and the doyen of the attribution and modeling arena, concluded marketing models do not have to be blunt instruments.  Data can and should be leveraged at diverse levels.  Yet untapped data has potential.  The people and the process must be right. 

Next story loading loading..