Most papers on digital advertising focus on the point of view of Internet companies such as Google and Microsoft, and were written by people working for those companies. Even though that might help understand better how such companies manage the ad auctions, advisers do not have access to the required data and cannot use those works directly to improve the effectiveness of their digital campaigns. In this paper, we describe a model for determining which keyword positions to aim for on search engines using users' navigation histories accessible to advertisers. The navigation history of users who interacted with an element of a campaign is represented by a flow in a graph. We use formulas that link changes in keyword positions with the number of clicks on the ads which appearance was triggered by those keywords, and show how these changes modify the flow. Taking into account budget constraints, we then describe algorithms that reorganize the flow by changing keyword positions in order to maximize the expected profit from conversions tracked online.
Published April 2017 , 17 pages