Marketers have a lot of misconceptions about marketing mix modeling (MMM).
But it’s far less expensive – and easier – to execute than it used to be.
Deli meat brand Land O’Frost has been investing in MMM for years, seeking insights into how different cultural trends, campaign strategies and channels impact its KPIs.
Historically, it’s been challenging to find “solid data points” to prove that certain channels, like paid social and influencer marketing, are driving sales, Land O’Frost CCO Saverio Spontella told AdExchanger. Marketers typically track metrics like impressions, CPMs and engagement, because they’re easier to measure than sales and ROI.
In late 2022, Land O’Frost partnered with research and analytics consultancy Big Chalk for help improving its approach to MMM and to figure out which opportunities it might have been overlooking – or even inhibiting – by not measuring ROI effectively.
The meat of it
Before partnering with Big Chalk, Land O’Frost was working with another MMM vendor, but made the switch because the brand wanted to take a nimbler, more streamlined approach to measurement, according to Spontella.
For example, advertisers have to understand which parts of their marketing mix led to impressions and how that relates to sales. They need concrete data they can use to justify their spending to colleagues – “like to my CFO,” Spontella said.
Determining exactly how each part of a marketer’s strategy impacts sales and impressions is where MMM providers come into play. “It’s not as complicated as it probably seems when you’re looking at MMM from the outside,” said Rick Miller, a partner and head of marketing effectiveness at Big Chalk.
In simple terms, MMM models estimate “how much volume of a product is sold when a certain tactic is in play at a certain time,” said Miller.
But despite MMM getting easier to use, it’s far from simple, and many marketers still make mistakes when applying it. In working with brands, Big Chalk has noticed several frequent pitfalls, which it highlighted in a report published on Monday.
AdExchanger Daily
Get our editors’ roundup delivered to your inbox every weekday.
Daily Roundup
One of the most common mistakes that marketers make, said Miller, is “misunderstanding the frequency with which you need to remodel.”
Lots of newer software touts its 24/7 capabilities and its ability to update daily or even hourly. But Miller warns against trigger-happy model refreshes.
Neither consumer habits nor media strategies change on a daily or weekly basis, so remodeling at that frequency is actually “counterproductive,” according to Big Chalk.
Start of something new
Instead, Big Chalk suggests using time series regression models – a fancy term for predicting future behavior based on historical data – divided by designated market areas over two to three years. This approach takes into account anything that impacts sales patterns, from weather and time of day to competitor brands running a promotion at the same time.
Big Chalk does analysis to show how much sales volume is driven by each marketing tactic along with its ROAS.
One of Land O’Frost’s biggest takeaways from its MMM results has been to redirect its dollars within TV advertising. As of just a few years ago, Land O’Frost was mainly focused on linear TV, Spontella said, and only “dabbling” in connected TV.
Based on Big Chalk’s findings, Land O’Frost saw that while linear was “moving a reasonable amount of volume,” Spontella said, it was extremely expensive – and a marketing tactic is only as worthwhile as the returns it rakes in.
It’s not that linear was “a bad tactic,” said Miller, it just “wasn’t the right price point.”
Too much of a good thing
But while some MMM recommendations are intuitive, like shifting from linear to CTV, other best practices are less obvious.
For instance, marketing mix models typically account for at least two years’ worth of data, so if you refresh a model after one month, that’s only four weeks of new data compared to the previous model.
Another common issue is that many marketing mix providers rely on geographic data that lacks sufficient granularity, limiting the model’s ability to deliver actionable insights. Segregating mix models by designated market area provides better comparisons between various markets, especially when a brand is executing media differently region by region, Miller said.
The biggest mistake marketers make when it comes to MMM, however, is not using it at all. Often, they assume it’s too expensive, too time consuming or both. But MMM has become both faster and cheaper over the past decade. Whereas it used to cost hundreds of thousands of dollars and take up to six months to execute, Miller said, now it can be done in less than two months, for under $100,000.
But although MMM is “easier, faster [and] cheaper now,” he said, it still needs to be done with a partner who understands the modeling science.
Otherwise, Miller said, the “mistakes can be really, really costly.”
