Ads Selection At Twitter

Online advertising is a multi-billion dollar industry and it also serves as the major revenue source for Twitter Inc. In this talk, we present the ads selection pipeline at Twitter, using Promoted Tweets in Home Timelines as an example. The pipeline starts from targeting, where we model Twitter users' attributes offline, e.g. user gender, age, interest etc, so that we can match them with advertisers' specified audience criteria. The second critical component is user engagement rate prediction, where we employ a large-scale online learning system to do real-time training and prediction with rich features. Lastly, we run a second price auction based on the predictions, advertisers' bids and some other optimization parameters. We will present a series of case studies drawn from recent experiments in the setting of the deployed system used at Twitter.