How to bring your marketing mix modelling into the 21st century

1. Evaluate your media by geography and market

Big TV buys may run nationally, but digital ads deliver on demand locally and offer granular reporting. Collect and model your data by market to help you get additional data points that make your model more representative.

Jeff Shatz, VP of marketing effectiveness at Nielsen, explains why this is critical. “The more granular the data, the more variability the statistical model is able to pick up, and the greater ability to tease out true drivers. If only national-level marketing data is used, the model will not be able to account for critical market-level influences that impact whether a purchase is made.”

2. Differentiate video platforms

Ditch the generic digital buckets that treat all online video impressions equally. Instead, break out video by platform so your model provides an individual read on each. Qualities like watch time, audibility and viewability vary widely across video platforms and, depending on what you’re trying to achieve, will have a different impact on the effectiveness of your media.

Recent research that we commissioned with Nielsen supports this, finding that when consumer packaged goods companies’ MMMs evaluated video platforms independently rather than aggregated, return on ad spend (ROAS) varied by as much as 48%.¹

3. Consider the various elements of your media plan

Assess different elements of your media plan by channel, including ad formats, audience segments and your campaign’s reach and frequency. From there, you can find stronger connections between your online efforts and your offline sales.

The Hershey’s marketing team does this by asking media partners and their media agency to supply data directly to their marketing mix modelling measurement partner, helping ensure the data is accurately broken out by brand and ad format. On YouTube, this level of granularity has helped the team establish that a specific mix of YouTube TrueView, Google Preferred and six-second bumper ads was most effective in driving sales for Reese’s Peanut Butter Cups. As measured by marketing mix modelling, the retail ROI for The Hershey Company's portfolio of brands increased 40% year over year from 2017 to 2018.

Clorox also took this type of granular approach, splitting out YouTube TrueView, Google Preferred, and six-second bumper ads, before sharing indexed ROI data with Google. An analysis of this data revealed that it made sense for the brand to move more of its YouTube budget into bumpers and TrueView ads to better balance its media mix and maximise ROI.

4. Validate what you learn through experiments

MMMs are inherently complex models – and they’re not perfect. Before making major changes to your media strategy based on marketing mix modelling results, run sales lift tests or other isolated experiments to test the impact of a single change in strategy.

For example, Frito-Lay North America, a division of PepsiCo, uses MMM along with sales lift and brand lift studies so it can evaluate three points of measurement before making any changes to its current media strategy. The results of all three methodologies will not always align, but this validation helps guide future media planning and justifies increased investments in a platform or strategy over time.

Frito-Lay North America has used three years of these lessons to hone its current YouTube strategy. Now the brand runs a specific mix of ad formats and maximises its best performing audience segments to optimise the efficiency of its YouTube investment. James Clarke, senior director of portfolio media, analytics and CRM at PepsiCo, says transparency has been critical to the brand’s success.

“We share MMM performance results back with Google so that both teams are aligned with and operating toward the same goals, instead of keeping those cards close to our chest, and limiting the capabilities of the partnership,” he says, reiterating the importance of sharing results with relevant partners.

5. Test your creative tactics

There are infinite ways to tell your brand story on digital. That’s why the most savvy marketers are taking the guesswork out of these decisions by using marketing mix modelling to measure the effectiveness of their creative.

Frito-Lay North America uses MMM – complemented by isolated experiments – to test the impact of the more personalised creative it has developed for YouTube. For example, as its marketing team leans into solutions like YouTube Director Mix, a tool that allows them to create customised videos at scale, they are seeing indications in their MMM results that more personalised creative is driving higher incremental sales than a message designed to have broad appeal.

“We want to provide consumers with the most relevant content based on what makes them tick,” Clarke says. “We’ve seen some early signals that suggest tools like YouTube Director Mix allow us to get closer to that goal without having a major impact on cost of media.”

Clarke adds that as they measure the impact of creative elements on sales, there have been some surprises. “Sometimes, creative elements we thought would be meaningful don’t have an impact; other times, things that seemed trivial really move the needle.”

Uncover what’s really working in your digital media strategy

With the flexibility and customisation that digital marketing offers, traditional measurement methods have to evolve. Setting up the right data inputs in your marketing mix model at the outset of your campaigns means more powerful, actionable measurement in the output. Ensure you evaluate your campaigns at the right level of granularity to get the most out of your modelling – and your media investment.