Display Advertising In Mobile

Kamakshi Sivaramakrishnan, Google, Mountain View, CA

Stevenson Hall 1300
12:00 PM - 12:50 PM

Modern smart phones and mobile web browsers display advertisements on consumer properties like the iPhone/Android apps and mobile versions of popular content publishers like m.espn.com, m.cnn.com, m.nytimes.com, etc. Similar to traditional ad-supported online web content model, mobile web content is also supported by advertisements as revenue models. Such a revenue model relies on market mechanisms to elicit prices for these advertisements, making use of an auction among advertisers who bid in order to have their ads shown. These advertisers, depending on the context of either web or mobile, target either keywords, content and mobile traffic profile respectively. The ad placement problem involves three parties--advertisers, publishers, and auctioneer (or ad network)--we present cutting edge algorithmic techniques that jointly optimize the goal for each of these parties. A commonly used business model in digital ad placements is pay-per-click where the advertiser participates in the auction for an ad impression by expressing a value for a possible click on the impression. As a result, the auctioneer actively tries to solve a critical problem, i.e., to predict the probability of click-through, commonly known as click-through rate. The algorithms for click-through rate prediction, traffic allocation and pricing use techniques from three mathematical areas: mechanism design, optimization, and statistical estimation. We present an overview of the statistical techniques that estimate the competitiveness of an ad and hence traffic it would receive when participating in the auction. The discussions will be generalized to the problem of digital advertising with topical references to the mobile advertising context at relevant points through the talk.