What’s in a name? MMM, multi-touch attribution and cross-channel attribution compared
Attribution invokes a ton of buzzwords, and navigating them can be tricky. Columnist Alison Lohse walks you through the differences between marketing mix modeling, multi-touch attribution, and cross-channel attribution.
Marketing analytics is one of the most buzzed-about categories in the current red-hot martech landscape. For those of us in the industry, it’s exciting to see so much interest and innovation around how we measure the effects of our efforts. But where there’s buzz, buzzwords will follow, and the marketing analytics space is no exception.
Even in my small corner of the world, attribution, the language used to describe methodologies and technologies is often murky. “Attribution” means different things to different people, especially across our dynamic ecosystem of vendors, agencies and brands, not to mention the ad tech side of things. It also gets jumbled up with other sophisticated measurement techniques like marketing/media mix modeling (MMM) and further complicated with data science lingo, like AI and machine learning.
Untangling all of those threads would make for a very long article, so let’s start small. What’s the difference between marketing mix modeling, traditional multi-touch attribution and today’s cross-channel attribution?
Marketing mix modeling
At a high level, marketing mix modeling (referred to interchangeably with media mix modeling, both shortened as MMM) works with top-down, macro-level information. MMM analyzes aggregated historical data, customarily from offline sources like TV, radio and print, and delivers organization-level metrics around planning, spending and performance. This analysis includes both internal and external factors, like marketing and pricing data, market elasticity, seasonality, and weather and news events (something like an election or hurricane, for instance, would play into an MMM evaluation.) It is usually performed once or twice a year.
Traditional multi-touch attribution (MTA) tackles marketing measurement from the other direction, taking a bottom-up, granular, user-centric approach. MTA looks at individual users’ digital journeys and analyzes the path to conversion across, as its name suggests, multiple touch points. This allows it to measure the impact of each individual marketing tactic.
A largely digital practice, multi-touch attribution uses data science to turn account-specific, near-real-time data into insights about how marketing initiatives are performing at each level, from vendors to channels to creative.
Both of these approaches offer pros and cons. MMM delivers the breadth of a 30,000-foot view into variables both inside and outside of the marketer’s control. As such, it’s valuable for metrics like the financial value of brand ads, for which a longer time frame and greater context are useful.
That longer time frame is also a drawback, however, impeding your ability to respond to circumstances in a timely way and optimize accordingly. Aggregate data also struggles to show the nuances of the customer journey, making it difficult to drill down and target specific audiences and strategies.
Multi-touch attribution has speed on its side, enabling marketers to understand and react to what’s happening on a daily basis. MTA’s granularity supports a deeper understanding of the synergies between factors as well, so you can make adjustments to placements and spend based on a holistic view into the performance of each step of the campaign. But its emphasis on digital is both a blessing and a curse.
While its digital roots give MTA the data it needs to dive deep into the customer journey, they can also lead to under-attribution of offline activities and a failure to take baseline conversions (those that would happen without any marketing efforts) into account. Finally, while traditional multi-touch attribution models improved upon first-generation last-click attribution, they still often rely on a single algorithm, which by definition only analyzes the data in one predefined way.
At the most basic level, MMM grew from and excels at measuring offline activity, while multi-touch attribution focuses largely on digital, online efforts. Each discipline will tell you that it now addresses the other (MMM claiming it can handle digital, for instance) but that’s typically true only to a limited point.
For this reason, many marketers run both analyses at different times for different reasons. Yet even then, there’s a gap; if you can’t integrate online and offline, you can’t truly understand how your overall marketing strategy is working.
Cross-channel attribution (not to be confused with its fellow hyphenate, multi-touch attribution) arose to meet this need. It brings together online and offline activities, high-level modeling and micro-level data, historical aggregates with data science. Cross-channel attribution bridges the gap between MMM’s strategic insights and MTA’s more tactical recommendations, allowing for both long- and short-term planning and projections.
How does it work? Our version of cross-channel attribution uses time as the common language between online and offline channels. With time as a common metric, we carve out a pie that accounts for digital, TV and radio, as well as baseline conversions. With time as a commonality, the MMM and MTA models can talk to each other and give us the big picture.
Cross-channel attribution can also give you the ability to dig into each piece of the pie, using the most appropriate model and algorithm, to get into the granular details of each channel. As a result, you can see both combined impact of offline and online channels and channel synergies — how all of the various channels work together toward a conversion goal.
Marketing analytics is evolving at a fast, fascinating pace, and attribution is no exception. It’s incredibly exciting that we no longer have to cobble together our offline and online metrics into a vague set of assumptions about holistic performance. MMM and multi-touch attribution still have their place in the marketer’s toolbox, but now that we can bring them together, why wouldn’t we?