Location-Aware Ad Recommendation to Bid for Impressions

We present our work-in-progress design and development of a location-aware advertisement recommendation system. The ad-delivery platform acts as a combination of a DSP (Demand Side Platform) and an ad exchange. Advertisers act as clients and set up ad-campaigns defining the ad-targeting and budget criterion. The live system connects to SSP (Supply Side Platform) and receives ad bid requests selling ad impressions. The bid request contains information about the device, user, location, impression type, publisher site info, etc. In response to a bid request, the ad delivery platform can decide to bid or not bid for an impression. Upon a successful bid, an impression is show which the user may click or ignore. This process follows an established protocol, the RTB (realtime bidding) and the entire process completes within 100s of milli-seconds. The objective of the ad recommender is to pick the most relevant ad that should be shown to the user and this ad form the part of the BidResponse, in case the ad-platform decides to bid for the impression. They key context that we use to pick the recommended ad is user's location. In addition to designing a location-aware recommender, we also describe our approach for identify interesting locations from raw gps trace in the impression bid requests received from SSPs.