AdWord Auctions: Fairness Without Loss

We introduce a notion of fairness in the models for sponsored search auctions that provides a chance to new advertisers, as well as, lower budget advertisers to obtain better slots. This enables the search engines (i.e., auctioneers) to estimate the quality or relevance scores for new advertisers, and to better cater to the advertisers in the tail, and thereby encourage more advertisers to participate. We show that such a fairness policy can be incorporated without any significant loss in revenue for the auctioneer, and potentially a gain in the long term, as more players find it profitable and easier to join the game. Additionally, the social value, i.e., efficiency, as measured by the combined value of the auctioneer’s profit and the payoffs of all the bidders, might actually increase. We study revenue and incentive properties of mechanisms for implementing fairness, and justify the above statements.