Experimental Designs and Estimation for Online Display Advertising Attribution in Marketplaces
Display-related advertising comprises over a quarter of online advertising revenues, driving 87.8% of media buyers to use real-time bidding to buy digital advertising space. As a result, attribution becomes a key player in determining where and how marketers input their advertising efforts. While randomized experiments have become the standard approach to attribution, they are far from perfect, suffering from multiple confounding effects. This article discusses common traps that could bias your results when running an experiment to determine attribution. Read to learn why the absence of a placebo in your control group may be the best option; explore the unintended consequences of your experiment in the marketplace; and discover the importance of defining a user selection policy, irrespective of whether you're using a CPA or CPM strategy.