Media Mix Modeling Or Incrementality Measurement: Which Will Replace Multi-Touch Attribution?
CTO and Co-Founder at Measured, helping brands grow by informing cross-channel media investment decisions with incrementality measurement.
Multi-touch attribution (MTA) has officially gone the way of the dinosaur. As data privacy restrictions like Apple’s App Tracking Transparency (ATT) and Google’s elimination of cookies inhibit the ability to track data at the user level, attribution methods anchored on building click paths are no longer viable. Although it was never smart business to rely solely on ad platforms to grade their own homework, the last-click attribution systems they use to report campaign performance are also suffering from the same privacy-driven accuracy issues.
So, what’s next? Businesses looking for measurement alternatives are increasingly being drawn to media mix modeling and incrementality measurement to replace outdated attribution methods. Both approaches control for a broad set of media and nonmedia factors that impact the consumer decision-making process. However, they’re not the same, and which approach to use is not an either/or decision—in fact, the best answer is “both.”
The Pros And Cons Of MMM
Media mix modeling (MMM), also known as marketing mix modeling, is a top-down approach that aggregates years of historical data and includes external influences like the economy, weather, and competitor activity. Then, multivariate regression models are used to predict the contribution of variables like marketing tactics and spend to business outcomes. As a result, marketers can predict how a different marketing mix could impact conversions and sales.
MMM does several things well. For established brands with years of data, MMM can effectively run long-tail effect, multilevel, and multistage models. It incorporates both non-addressable media and nonmedia variables well and delivers meaningful analysis at a high level. It’s especially good for enabling a strong understanding of how marketing and nonmarketing variables impact sales, as well as forecasting and budgeting for different potential scenarios.
But when marketers need ongoing analysis or insights at a tactical level, MMM is unable to deliver. Due to a lack of granularity, it can’t offer insights into campaign-level performance and decision-making. MMM also struggles in the face of unexpected market disruption (hello, Covid-19, and iOS 14.5) when there’s no data from a comparable period in the past. It only knows what it knows.
Brands looking to build out MMM capabilities face the classic build-or-buy dilemma as well. Many of the top MMM providers require significant budgets—a result of their highly consultative approaches. However, building internally requires both in-house data expertise as well as a significant effort to ingest, clean and automate data across a wide array of sources to produce repeatable results. Manual data processes and the inclusion of nonmedia variables that are unavailable in near time can lead to results that are updated infrequently (typically, quarterly) and often lag in-market outcomes by up to 30 days.
How Incrementality Measurement Stacks Up To MMM
Incrementality measurement can address many of MMM’s shortcomings and, used concurrently, provides a truly comprehensive picture of marketing performance. But first, what exactly is incrementality measurement? At the highest level, incrementality measurement segments audiences into statistically significant test and control cohorts and withholds the media being tested from one of the audience groups. The difference in conversion rates between the two cohorts can effectively reveal incrementality: the true incremental contribution of a media channel, campaign, ad set or tactic.
Although MMM lends itself to big strategic portfolio considerations and long-term planning cycles, incrementality measurement helps marketers make critical ongoing decisions about channel spend and campaign management. It offers daily reports instead of quarterly ones, leveraging automation to return results in close to real-time. It can also deliver fast, cross-channel, source-of-truth reporting for every marketing channel, so businesses can respond quickly and effectively.
But because of the more tactical focus, incrementality measurement is best for ongoing campaign and spend optimization, not long-term strategic planning. Incrementality experiment design and execution can also get very complex, very fast. On the surface, running a holdout test may appear quite simple, but each channel has unique nuances in technology and capabilities. So, advanced data science is often required to create scientifically significant control groups and achieve clean reads, and continuously ingesting and normalizing data from sources across the business is no small task at any level.
Fortunately, it’s becoming more accessible for midsized businesses to use both MMM and incrementality. MMM was historically considered a tool that only enterprise-level companies could afford, but advancements and automation have made it easier to take on. Using incrementality to address the gaps of MMM makes that investment even more valuable, especially because if a brand is set up for one or the other, it’s likely well-positioned to deploy both.
Taken together, incrementality can validate and enhance MMM. Its experimental results help avoid correlation/causality issues by using incrementality outputs to train the MMM model. Incrementality provides fresh data on which to make rapid tactical decisions, complementing MMM’s long-term strategic and planning strengths.
Should MMM or incrementality measurement replace MTA and platform-level reporting? The answer is yes and yes. Moving forward, the combination of MMM and incrementality will likely be the gold standard in a post-iOS, post-cookie, post-(insert next policy here) world.