Netflix built its DVDs by mail business by combining a convenient hassle-free movie service with a movie recommendation system. The original recommendation system, Cinematch, which relied on movie rental history and user ratings of movies, was somewhat limited due to a lack of data. It was modestly successful, however, and helped Netflix establish itself as a leading online movie rental house. Netflix always knew that Cinematch needed to improve and attempted a crowdsourced competition in 2006 to improve the system. The contest ended three years later with a 10% improvement in movie recommendations. Since then Netflix has dramatically expanded its online streaming service, which allows it to capture a whole new set of data points on users' actual viewing of movies. For example, did a user watch the whole movie or did they stop it after 5 minutes? The expanded data set improves Cinematch's accuracy. Improving Cinematch's accuracy is crucial to Netflix's future, not only to compete with other online streaming services, but because a smart recommendation system can direct users to content Netflix already owns thereby limiting content licensing fees.
Read the article.
(submitted by Walter)