When the number of movies and series on Netflix increased exponentially, the CEO of Netflix, Reed Hastings, realized they needed to improve their recommendation system to retain customers. After Netflix made several unsuccessful attempts to develop a new algorithm to improve their movie predictions accuracy, they decided to start a three–year open contest. When they launched the challenge, they invited their lead users to come up with a better recommendation system than the one Netflix had at that time, named Cinematch. Many mathematicians, statisticians, software engineers and cyber geeks, symmetrically skilled, from all over the world participated. The variety of participants is often beneficial since somebody outside the field of the problem can often come up with a good solution (Harvard Magazine, 2013). To win the contest, the proposed system had to improve recommendation accuracy with at least 10%, compared to Cinematch. The ones who would achieve this goal, received a price of one million dollars.

Participants received a data set of 480.198 customers, who rated 17.770 Netflix movies, where personal information was removed (Thompson, 2008). Then they had to develop algorithms to improve the movie prediction accuracy. Especially the movie Napoleon’s Dynamite, an indie comedy, seem to be extremely difficult to predict for customer preference as it was a culturally and politically polarizing movie. In the header you can see a picture of Bertoni and his son, attacking the “Napoleon Dynamite” problem (Thompson, 2008). More than 44.000 submissions, spread over 186 countries were sent for the challenge and afterwards, the received preference predicting algorithms were compared to the movies customers actually picked.

Chris Volinsky (left) with Robert Bell working out Netflix algorithms at the AT&T research labs in New Jersey (Thompson, 2008)

After three years, BellKor’s Pragmatic Chaos, a multinational seven-person team got an increase of more than 10%. The team consisted of people that used to compete with each other. When they started cooperating, they worked from different locations and communicated namely via email. The first time they met was at the prize ceremony. At their speech they said that because of all these different algorithms they created, they could combine them to create an even more complex one. For example, some examined movies in bundles of element (genre, actor), where others looked at what movies were rated rather than how they were rated. The team manager, Chris Volinsky said about their strategy: ‘You need to think outside the box, and the only way to do that is finds someone else’s box’.

Netflix said the new algorithm increased their customer retention rate and they admitted Netflix obtained their brand new search engine for a bargain price. As the contest was so successful, they decided to launch other open contests for shorter time spans of 6 to 18 months.

The way teams came together, especially in the end, suggests this way of collaboration via the internet can be used for other complex issues as well. This shows challenges in science, technology and business can be served by the wisdom of the crowd.

By 363966lk

Bibliography
Ideaconnection (n.d.), Open Innovation: Netflix Price, Idea connection, https://www.ideaconnection.com/open-innovation-success/Open-Innovation-Netflix-Prize-00032.html retrieved on 13-2-2016

Harvard magazine (2013) More Shots on Goal, http://harvardmagazine.com/2013/12/more-shots-on-goal retrieved on 13-2-2016

Thompson, C. (2008) If You Liked This, You’re Sure to Love That, The new York Times, http://www.nytimes.com/2008/11/23/magazine/23Netflix-t.html?pagewanted=all&_r=1 retrieved on 13-2-2016