The contest

The plista Contest is the search for the best live recommender algorithm.

The contest is an open competition for the best live collaborative filtering algorithm to recommend relevant online ads as well as relevant editorial content based on previous user behaviour.

plista is one of the most successful advertising networks in the German-speaking world. We here at plista think that we have one of the world’s best live recommender algorithms. But we're quite interested in what else is out there...

The essential aim of this contest is to improve the performance for On-Site-Recommendations, which is our name for the little widgets that can be found on many news and other high-traffic sites. These widgets recommend to the visitor other items that may be of interest to them. They are often titled: "You might also enjoy reading..." or "Customers that looked at this item have also looked at..."

Here is an example:
Alice is reading an article about politics on a news website. Which other articles from that same website are of highest interest to her? One solution is to simply randomly recommend more articles from the same category to her. Sometimes this works out and she actually clicks on the recommended links, but this random approach is of course not the most viable one, there are many factors that may influence the choice of recommended items, for example:

• the time of day
• the geographical location of the visitor
• the item the user is currently viewing
• etc.

The goal of this contest is to find the best performing algorithm, where performance means the actual success rate (measured in clicks per times displayed) but also the performance under time aspects. This last point is critical: we are talking about hundreds of requests per second from several million users and 100,000s of items to select from.

Please read the rules and sign up if you're interested in participating.