What are the most effective attribution models? A team at MarketShare set out to answer this critical question with a major big data study.
The study simulated over 300 million online shoppers, “exposed” them to ads, and recorded the ways the “shoppers” did (or did not) respond to the advertising they received. Then, the team analyzed how well various attribution models fared at assessing the influence of those ads. The team did this by comparing the attribution that each model suggested with the actual influence of ads, as recorded by the simulation. The goal was to see how closely each attribution model came to describing the “reality”—and thus to seeing how effective each attribution model really is.
Read this brief to understand how well first click, last click, matched pairs and discrete choice models—the predominant models for digital attribution—fared when held to the scrutiny of big data.
(Then, learn how well each of these models performed at accounting for customers who seem to respond to ads—but in reality would have made a purchase with no help from the ad at all—in the separate MarketShare brief: “When is Advertising Unnecessary? - Attribution and Customer Intent.”)
To find out which models were most effective, download the brief now.