Commentary

'Marketing Mix Modeling: Alive And Revitalized'

  • by , Op-Ed Contributor, November 17, 2022

So said Sequent Partners’ Jim Spaeth, summing up the third day of the 2022 “Attribution & Analytics Accelerator” conference, co-hosted with the Advertising Research Foundation.

“[Marketing-mix modeling] can now support a broader scope of more granular, quality findings based on its comprehensive view of the full marketing funnel,” he added, echoing many of Wednesday’s speakers.

Spaeth suggested practitioners take care in interpreting and adopting the optimizations, re-targeting, and/or predictions of any model.  

The challenge of disparate, fragmented, and non-equivalized data as the inputs to marketing models was highlighted exquisitely by In4mation Insights Partner and Co-Founder Mark Garratt.

He underlined that the key in modeling is estimating coefficients for very different levels of data inputs within the same overall analysis.

As an example, data for retail stores that have “feature and display” investments versus those that do not, can lead to what he called “data sparsity and mismatch.”  

“Mathematical techniques matter,” he said.

According to OLLY Insights Director Liz Riley, Garratt’s “hybridized” modeling approach works for her nutrition brand.

PepsiCo Director of Global Media Analytics & Digitalization Sameer Kothari said his company has reinvented the use of marketing-mix modeling and developed a proprietary system called “ROI Engine.”  

Working with Middlegame, PepsiCo has developed a system to predict in-flight ROI outcomes of digital campaigns based on attention, creative and audience mix.

This digital, real-time optimization engine allows the PepsiCo team to dissect campaign-level incrementality into media efficiency, creative and targeting factors of its digital executions.  

Like most presenters over the three-day event, how PepsiCo defines media impressions in its model inputs would be revealing in view of the fundamental requirement for measurement of a person’s attention – or “eyes-on” – to an ad in order to drive a brand’s outcome.  

Kothari importantly acknowledged the value of different modeling techniques, depending on the category, environment and value of short-term versus long-term considerations.  

Keen CEO Greg Nolan reviewed the speed, agility, and flexibility of enhanced data sourced marketing mix models that incorporate machine learning.  Specifically, he said this approach enables companies to cope with any marketing pivot needed to maintain profitability, volume, or new product support goals due to major disruptions at any time.

Stay tuned for my observations of the fourth and final day of the conference in my op-ed tomorrow.

Next story loading loading..