Flaws in Attribution Are Costing the Industry Billions in Profits - 5 Mistakes Every Marketer Should Avoid

A revelation has recently come to light– that multi-touch attribution (MTA) models have fatal flaws that are costing companies billions in sales and profits.

An informal consortium of companies came together to look into the matter. The consortium included a few marketers, including a prominent insurance company, a leading advertising software company, a couple ad agencies, and a few academics. Each had independently found fatal flaws in the recommendations coming out of popular attribution solutions. These budget re-allocations were so contrary to effective spending that, when applied, they actually reduce the amount of sales per dollar of marketing spend. One retailer reported faithfully applying the findings of the MTA for two-years, yet, business growth didn’t materialize.

Some questioned if the attribution vendors owned by media companies were self-serving in their inaccuracy. They suggested either fraud or system gaming. Others were more deferential, noting that applying math to a problem can create an illusion of good science, but may blur obvious logic. They pointed to over-reporting the value of impressions delivered to people that would have bought anyway. Whatever the reason for the flaws in attribution recommendations, the consortium members all believe in the value of analytics and want to ensure the industry gets the analytics right. As one member explained, “The consortium’s desire is to shine a bright light into these dark, black boxes of attribution. This will help the industry make better decisions to create more value.”

The purpose of this brief paper is to outline the findings from the consortium and to open a dialogue in the industry so we can ensure we get the science of attribution right. There are billions at stake, and we have obligation to the professionalism of the industry to identify problems and promote best practices. The remainder of this paper is organized into two sections: First, a review of three independent validation studies that identified the problems with attribution. Second, a discussion of the biases we identified, along with advice on how to avoid the attribution biases.