Advertiser Learning in an Internet Ad Network

Abstract

Do advertisers learn about the efficacy of their Internet display advertising over time? If so, does mitigating their initial uncertainty improve welfare for advertisers, publishers, and the ad network? We use data from a display ad network to estimate a model of advertiser demand. At first, advertisers don’t know how well their ads will perform at a new site. But after running ads, they see their click-through rates. If their click through rate is higher than expected, they are more likely to place future ads at that site. We use the model estimates to simulate how prices and ad placements would change if advertisers knew more from the start. We then assess the welfare implications of learning.

Date
Links