Netflix touts improved movie-ratings system
PHYSICAL: Prediction-system changes stem from contest
By Danny King -- Video Business, 5/11/2009
MAY 11 | PHYSICAL: Netflix has improved its Web site's user-generated movie-ratings system by changing how suggestions and rankings are factored into its predictions of how much a particular user will like a certain title.
The ratings improvements will be used largely in the "Movies You'll Love" section of Netflix's Web site, the company said on its community blog last week. The largest U.S. movie-rental service during the next few months will test how the new ratings system can be used by subscribers on their personalized Web sites, the company said.
"Members who rate movies will now receive better guidance on how many stars they’re likely to give movies they haven’t seen," wrote Todd Yellin and Jon Sanders, Netflix's director of product management and director of recommendation systems, respectively, on the company's community blog last week.
The improvements are part of a Netflix contest started in 2006. Users who submit suggestions that improve the company's Cinematch movie-ratings system are paid cash totaling $50,000 a year for a so-called "progress" prize and $1 million for a grand prize. The contest expires in October 2011.
The popularity of both Netflix—the company broadened its subscriber base by 25% in the first-quarter from a year earlier to 10.3 million—and computerized movie-ratings services have spawned unaffiliated Web sites such as Jinni.com. Such sites use algorithms to mathematically compute movie preferences for users who enter information about movies they like.
