The blunt edge of marketing mix modelling: Can it be sharpened for retail media’s rise?
Opinion
The ability of MMMs to pull together and scan diverse data sources to help show the impact of media investments on sales is not keeping pace. Can they be made fit for purpose?
Tools can get blunted over time; they lose their edge and ability to perform as initially envisaged.
This is becoming apparent with marketing mix modelling (MMM). The data-based forecasting methodology has become embedded in media budget planning over the decades. Still, we are now in a quite different environment from when the first computer-powered models appeared in the 1970s.
In our fast-moving landscape, trends appear and disappear overnight, sales can spike on the back of a viral meme, and media channels present new opportunities with eye-blinking rapidity.
The ability of MMMs to pull together and scan diverse data sources to help show the impact of media investments on sales is not keeping pace in a world where marketers and their agencies need to be able to analyse, predict, and act to meet real-time consumer behaviours and signals.
This begs the question: are they fit for purpose, and if not, can they be made so?
Time is of the essence
The matter is more pressing now that Retail Media Networks (RMNs) are moving onto centre stage and becoming a core channel – IAB figures show UK online retail media spend is projected to reach £1.5bn in H1 of this year. Marketers are increasingly looking to RMNs to access and leverage vast swathes of permissioned first-party data to optimise segmentation, targeting, and execution.
The lack of confidence in the ability of MMMs to handle cross-channel measurement and analysis within the RMN sphere is evident in a new report, The Future of Marketing Mix Models in the Age of Retail Media, produced by MediaLink and the Digital Shelf Institute.
It shows that only 35% of senior leaders are confident in their MMMs’ ability to measure incremental impact across channels. And in a world where automated investment decisions are made every minute, a majority (65%) receive new data and insights from their MMMs quarterly or even less frequently.
As Ash McMullen, head of ecommerce, Advantice Health, says in the report: “Last year, we embarked on a discovery journey with different vendors trying to find the magical MMM model. At the end of that process, what we found is that, for the investment level we have, it didn’t seem like the magical model that we were looking for existed.”
While Chandra DiGregorio, vice president of data science and research, 84.51° observes: “The MMMs do take a while. It feels like oftentimes you are seeing results maybe six months or nine months after you’ve run something, and by then the budgets are already set, so how is that moving the needle?”
Retail media is maturing as brands find it can help move consumers along the full funnel from awareness to conversion. The IAB Europe Attitudes to Retail Media Report shows that the number of brands working with four–six networks more than doubled, indicating recognition of the reach and scale RMNs offer. Now, MMMs need to evolve to properly integrate findings from these data-driven networks into their analytical frameworks.
What does this look like in practice?
Successful evolution requires grappling with several challenges. These include moving beyond aggregate-level analysis and taking on board granular attribution, digging in to isolate the true impact of retail media spend on sales, brand lift, and customer lifetime value, and signals at a shopper level, and integrating rich data sets to isolate the real impact of brand lift and customer lifetime value.
Reporting cycles also need to align much more dynamically with the agility of retail media campaigns and the ability to dial investment up or down, sometimes daily.
The full range of retail media needs to be fully grasped, from on and off-site to in-store, from programmatic to CTV integrations. It’s a complex ecosystem that requires a holistic understanding of how these touchpoints interact and can amplify effectiveness. Models must also be able to filter out the effect of operational levers, such as pricing and promotions, from results.
It’s a big challenge and relies on RMNs also getting their house in order, especially when it comes to measuring impact. Across the ecosystem, there is a lack of cohesive measurement standardisation and transparency, while even within individual retailers, there are too many siloes.
Ben Galvin, senior director of omnichannel retail sales and ecommerce at Monster Energy, outlined his frustrations: “When it comes to retail media, it is very important to realise that the effectiveness of it differs by retailer. It is really a whack-a-mole environment. You may have very effective tactics and measurements on Albertsons that are not as effective on Kroger, for example.”
MMMs can still be given the requisite tune-up to ensure they do not become obsolete, and machine learning technologies will be at the heart of the solution. But it’s not speed alone that will make MMMs useful.
Critically, models will need to become smarter and more sophisticated to make sense of a modern customer journey. If MMMs evolve, they will be a boon to brands seeking that critical reassurance before making their retail media investments.
Donna Sharp is MD at MediaLink, partner at United Talent Agency