Decoding attribution and marketing mix models: blog

Leith Higdon, director of analytics and insights at Mediabrands, with everything about the two that you wanted to ask but were too afraid to.

By: Leith Higdon, director of analytics and insights at Mediabrands

There are a few questions I’m hearing asked more and more frequently these days. Does attribution work? How do I measure media ROI? The answers are, in one way, plain and simple: attribution does work and media ROI is best measured through marketing mix models (MMMs).

But there appears to be confusion in the media industry about what attribution and marketing mix models are and how we should use them. Attribution and marketing mix models both assign credit for sales conversions (or other dependent variables such as traffic) to media, but that is where the similarities end.

To be clear: one is not a replacement for the other.

What are attribution and marketing mix modeling?

Any media measurement methodology should be selected based on the questions we want answered. Do we want to measure ROI or do we want to optimize our media mid-campaign?

The best way to think about these two media methodologies is in terms of top-down and bottom-up.leithHigdon

Marketing mix models are a top-down approach. They start with sales or another dependent variable and then look at all the things that could influence it such as media, distribution, weather, pricing, GDP, etc. MMMs explain which variables influence sales and by how much.

Attribution is a bottom-up approach. It starts with a consumer journey (usually digital) and tracks which ads a consumer sees or what a consumer searches for before they end up converting (usually online) in some way by purchasing a product, signing up for a newsletter or downloading a coupon. Attribution aggregates all the individual consumer journeys (and there are thousands of them) and looks at the consumers who purchased versus those who didn’t.

What are attribution and marketing mix modeling good for and why?

The 4A’s and the Coalition for Innovative Media Measurement recently commissioned Sequent Partners, a leading analytics consulting firm, to do an analysis of the MMM and attribution landscape. They found that the quality of methodologies and data used in MMMs are strong enough to consider it an ROI best practice. Moreover, they found that the level of data granularity available in the attribution methodology meant that it was better for mid-campaign optimization.

Marketing mix modeling has expanded from something mainly done by CPGs to all industries. It has a big advantage in its ability to incorporate those non-media variables (weather, the economy, price etc.) that could help sales.

This means that sales/conversions that get assigned to media in an MMM are a truer representation of what’s happening in the market compared to an attribution approach, which typically only accounts for a few variables. The credit that we give media in an MMM is accurate, meaning the return on investment is more accurately measured.

However, marketing mix models aren’t nimble enough for mid-campaign optimizations and are not designed to operate at a program or site level. That is where attribution comes in.

Attribution is still considered a new methodology. The number of approaches and the quality of data going into attribution models varies wildly. Most of these methodologies haven’t been around long enough for their quality to be fully tested, but they typically look at consumer journeys by linking two or three variables together – TV spots to web visits, digital impressions to online sales, and so on. Some might add in weather variables, but they typically exclude a good portion of the factors that are influencing conversions. This is why ROI from attribution isn’t as accurate as ROI from a marketing mix model.

Of course, the more variables added into an attribution model, the better it will be. That said, since attribution needs minute-by-minute data, it limits what can be included.

The methodology should match the question

While these two methodologies – attribution and marketing mix modeling – take different approaches, they should be used in complementary fashion to get a full picture of campaign effectiveness. In fact, for many advertisers, using both methodologies provides a clearer picture of the advertising ecosystem.

As the media industry is called upon more frequently to prove its effectiveness, it is in everyone’s best interest to not just become familiar with these two ways of measuring media, but to begin implementing both when the need arises. After all, two methodologies are better than one at proving how and when media works.