In 2009 Netflix held a contest challenging teams of developers to create an algorithm to increase the accuracy of its film recommendations by 10 percent. The award was $1 million, and it turns out Netflix pretty much flushed that money down the drain. The company doesn't use the winning algorithm. So, why?
The film rental/streaming service predicts how much a user might like a movie based on that user's ratings of other films, and Netflix has been improving it for years. Hence the contest, which offered different prizes, with The Netflix Prize being the grand prize.
The Progress Prize, a smaller one, was awarded to a team named Korbell in 2007 for creating an 8.43 percent improvement (which reportedly took 2,000 hours of work and a combination of 107 algorithms).
But it took another two years for a team to win The Netflix Prize. BellKor's Pragmatic Chaos, the winning team, managed to create a 10 percent improvement (and win the cool $1 million) in 2009, but its algorithm was never implemented.
The reason is two-fold, according to the Netflix blog.
First, the company claims the engineering effort and funding wouldn't be worth the results. Second, when Netflix infamously attempted to split its company into a DVD-by-mail service and a streaming service (and subsequently lost a sizable chunk of customers in the resulting fallout), and many subscribers canceled their mail service. To Netflix, this makes the rating system less important.
The blog states, "streaming members are looking for something great to watch right now; they can sample a few videos before settling on one, they can consume several in one session, and we can observe viewing statistics such as whether a video was watched fully or only partially."
Personally, this is difficult news to report, since I have honestly given up my own free will when it comes to movies. If Netflix thinks I'll give something four stars, then I watch it. And inevitably give it four stars.
Has it all been a lie?
Via Ars Technica