Abstract

In contextual advertising advertisers show ads to users so that they will click on them and eventually purchase a product. Optimizing this action sequence, called the conversion funnel, is the ultimate goal of advertising. Advertisers, however, often have very different sub-goals for their ads such as purchase, request for a quote, or simply a site visit. Often an improvement for one advertiser’s goal comes at the expense of others. A single ranking function must balance these different goals in order to make an efficient system for all advertisers. We propose a ranking method that globally balances the goals of all advertisers, while simultaneously improving overall performance. Our method has been shown to improve significantly over the baseline in online traffic at a major ad network.