Display Advertising's Impact: Biases and Temporal Variation in Effectiveness

Abstract

In this paper, we investigate ways to accurately assess and enhance display advertising effectiveness. We construct a two-month panel of browsing behavior data that covers nearly one million users, and over 24 million display ad impressions served on behalf of 400 advertisers. We discuss the importance of accounting for two types of selection biases: targeting rules and the consumer’s browsing intensity. When ignored, selection effects overstate display advertising response by more than five-fold, with browsing intensity accounting for one-third of this bias, and targeting accounting for the remaining two-thirds. Finally, we investigate the relative impact of entertainment and outcome oriented browsing on consumer response to display advertising, and examine whether such a relationship holds equally well for advertising campaigns that focus on early and late stages of the purchase cycle. We show that consumers are nearly twice as responsive when browsing in an entertainment oriented state, that a campaign targeting entertainment oriented users is expected to generate a 15.3% more conversions, and these effects are strongest in the early stages of the purchase cycle.