This is a guest post from the FT’s media correspondent, Kenneth Li
Finding a good movie is generally a hit or miss proposition, Reed Hastings, chief executive of Netflix, explained on Monday.
That’s a problem for Mr Hastings, whose billion dollar mail order movies rental business hinges on subscribers being progressively happier about what they watch next or risk dumping the service.
Netflix, which discovered early on that movies that were recommended generally pleased customers more than even new releases, spent a decade struggling to take the guesswork out of what to watch next and were unable to boost accuracy by 10 per cent.
Mr Hastings turned that challenge — telling customers what they want before they want it –- into a million dollar global challenge. The winners were announced today.
Recommendation engines underpin everything from consumer retail services such as Amazon.com to the “people you may know feature” on social networks such as Facebook and MySpace.
Three years ago, inspired by the British government’s 18th century challenge to solve a “vexing and dangerous aspects of transoceanic travel” in accurately plotting a ship’s longitude, Mr Hastings launched a million dollar challenge to see if smarter people from the world of engineering and computer science scientists could do better. (The movie Longitude is available on Netflix, Neil Hunt, chief product officer, cheerfully volunteered.)
“We’re trying to move it where nearly every movie you watch you love,” Mr Hastings told us shortly after announcing the winners of what has turned into the Super Bowl of the geeks.
Some 40,000 teams in 186 countries submitted solutions since 2006. Netflix provided 100m anonymous movie ratings with personal information stripped from the data set.
The winners, BellKor’s Pragmatic Chaos, was formed from the merging of three teams from Austria, Canada, Israel and the United States that combined their work to seal the prize. Sitting on stage in New York to accept the million dollar cheque was the first time they ever met.
Like the movies, this battle has a sequel. Netflix announced a second million dollar challenge on Tuesday. The latest contest challenges researchers to come up with a better service by sifting through more granular data including subscriber age, gender, zip code, but from a pool of customers who rarely submit ratings. The information will be provided anonymously and untraceable to actual customers.
