Paid Search: Modeling Rank Dependent Behavior

Using disaggregated data from a Chinese search engine we jointly model ad rank and performance for hospitality related keyword searches. As a result of our modeling framework we can better determine the optimal keyword bidding strategy for an advertiser given the search engine´s control over ad rank. Our approach removes rank bias in estimating keyword bidding performance. We then illustrate the impact of branded versus generic keyword searches, outlining profit maximizing keyword bidding.