Netflix Recommendation Algorithm Directs 75% of Viewership
According to a tech blog post from Netflix, the company’s recommendation engine drives 75% of content viewership for the service.
In 2006, the company announced its Netflix Prize, a machine learning and data mining competition regarding movie rating prediction. Netflix offered a $1 million prize to see who could come up with the best recommendation algorithm, and BellKor’s Pragmatic Chaos team went on to win, with a 10% improvement over Cinematch on the test subset. Since the contest came to a close, Netflix now asserts that “everything is a recommendation,” from a straightforward Top 10, to unorthodox, super-specific genres, like “Imaginative Time Travel Movies From The 1980’s.”
When Netflix displays content under a recommended genre, the movie picks are personalized three times, “the choice of genre itself, the subset of titles selected within that genre, and the ranking of those titles.” Basically, the algorithm will take into account the diversity of customer’s household, and account for different tastes of various members, along with their possible mood based on previous viewing choices. The system also caters to a user’s awareness – Netflix offers relative explanations to its recommendations, prompting the viewers to trust the service. Finally, similarity, freshness and current movie ranking are also taken into account in recommendations.
With the focus of the recommendation engine being so personalized, it’s not hard to believe that the service makes up large part of the general Netflix viewership.