A short history lesson in the development of Media Mix Modeling (MMM) and Multi-Touch Attribution (MTA) may be all one needs to understand their essential differences.
Traditional MMM – a mature, decades-old solution, founded by economists – is backed by an economic theory that substantiates causation behind the models.
MTA, on the other hand, arose from a technology-centric point of view that answered, “how do we handle this size of data”, versus “what is the economic underpinning behind this digital data and what’s the most appropriate way to model it out”.
These discrepancies in history between MMM & MTA have led to differences in the way data is classified and models are deployed. In MTA, there is little consensus on the correct way to model. Transparency in process, as a result, is often lacking largely because the majority of approaches are not clearly understood. Often, we are left with incomplete or unusable results.
The way to model a digital click-stream is still up for debate. In fact, eMarketer’s recent report showed that fewer than 10% of US marketers think their company’s attribution knowledge is excellent. We clearly have a long way still to go. In the report, Mike Bregman, SVP of data & analytics at 306i states “If you are trying to grow the business, you should begin with a top-down methodology. From there, you should think about content optimization and how much to invest channel-by-channel.”
At Truesight, we believe that digital modeling comes down to classification. We must understand why people bought or didn’t buy and in that understanding we must control for the underlying time series impact of other macro drivers.
A comprehensive, dynamic top-down model that blends macro drivers with individual drivers is perhaps today’s gold standard solution. Given the available data, a bottoms-up only model for attribution is unrealistic as a standalone solution for MTA.