Netflix, Dark Knowledge, and Why Simpler Can Be Better

White Paper


In October 2006 Netflix unleashed its challenge to crowd source one of its most difficult challenges: making quality movie recommendations to its customer base. The competition goal was to use a customers’ previous movie reviews to predict how many stars they would give a particular movie in the future. $1 Million was promised to any team able to beat Netflix’s current algorithm by more than 10% in root mean squared error.

Despite the success of the competition for Netflix – both technical and marketing – they never used the winning algorithm. Netflix had paid out a million dollars for this product – why not use it?

Download the White Paper