On the interest of introducing randomness in ad-word auctions

Abstract

Search engines play and will still play a major role in the use of networks. Sponsored search auctions is the basic tool for a return on investment in this industry, accounting for an increasing part of the business. We introduce here a model for consumer behavior in the context of ad-word auctions. Considering that unsatisfying answers of the ad-word engine will lead some consumers to perform again the same request later on, we show that displaying only the highest bidding or highest revenue-producing advertisers in a deterministic way is not always the best strategy for the ad-word engine. Instead, some randomization among advertisers can provide higher revenues. We also design a Vickrey-Clarke-Groves auction rule for a display probability and compare it with the current generalized-second-price scheme.