It’s no surprise that Netflix has launched another contest to improve its movie-recommendations system—the $1 million the company gave away for the first Netflix Prize was the steal of the century. On Monday, three years after the contest kicked off, Netflix awarded the jackpot to BellKor’s Pragmatic Chaos, a team of seven engineers, mathematicians, and computer scientists who managed to improve the DVD-rental service’s recommendations by 10 percent. BellKor beat out another team— the Ensemble, with more than 30 members—that had achieved the exact same ratings improvement but lost by turning in its method 20 minutes after BellKor.
Imagine if Netflix had paid all these math whizzes the prevailing wage—say, $100,000 a year. The company would have had to shell out more than $3 million for just one year of the top performers’ time, and that’s assuming it could’ve sussed out who the top performers were going to be. Of course, many of the programmers worked far longer than a year, some of them setting aside their primary occupations in order to work on the Netflix problem full-time. As Netflix CEO Reed Hastings admitted to the New York Times, “You look at the cumulative hours and you’re getting Ph.D.s for a dollar an hour.”
But even that number discounts the contest’s true benefits to Netflix. Had the company simply put out a help-wanted ad for software engineers, it probably wouldn’t have been able to recruit many of the geniuses it found through the competition. That’s because most of them already have other jobs. BellKor’s members work for, among others, AT&T and Yahoo, and many members of the Ensemble are employed by the data-consulting firm Opera Solutions. The participants also spanned the globe. Netflix got submissions from people in more than 100 countries, and the winning team’s members worked in New Jersey, Montreal, Israel, and Austria.
The Netflix Prize isn’t entirely novel. Science prizes date back to at least the 18th century, when the British government offered ₤20,000 to the first person to come up with a way to determine a ship’s longitude. Other notable innovation-prize winners include Charles Lindbergh, who won the Orteig Prize in 1927 for flying across the Atlantic, and SpaceShipOne, which won the 2004 Ansari X Prize, a contest to launch a privately funded manned craft into space.
But the Netflix effort was unusual for a couple of reasons. First, it was funded explicitly for the benefit of a private company; though many participants were interested in finding better ways to predict fickle human tastes, Netflix was looking for research that would help boost its bottom line. The company has already folded some of what it learned from the contest into its recommendations system, and that has helped increase its customer-retention rate. The Netflix prize is also notable for what it was after—not a feat of derring-do, like Lindbergh’s, or one of engineering, like SpaceShipOne, but rather a kind of mathematical recipe. Netflix was looking for an idea—and it turns out that Internet-enabled collaboration is particularly well-suited to fostering such abstractions.
Indeed, the Netflix Prize should serve as a model for other tech companies working on hard problems, as it combines the best parts of open-source development with the best parts of proprietary code. Just like an open-source project, the prize was remarkable for its spirit of cooperation—over the life of the contest, competitors frequently became collaborators, joining one another’s teams when they realized that they could never win by going it alone. The winning team is actually a mélange of three different teams; they joined up in June, and their winning algorithm combines the best ideas of each group.
Open-source projects work similarly, but they can sometimes become unwieldy and unfocused when they grow too large. What’s more, the open-source model can put off developers who—not unreasonably—are interested in some kind of reward for their work. The prize model solves those problems. Because there’s a reward involved (not just money but also fame), teams have a natural incentive to stay focused on a goal and to closely monitor each participant’s progress. The winning team—like several others that participated—worked out a formula to determine each member’s share of the prize money. And, of course, the prize model works out much better for a sponsoring company like Netflix. By awarding a prize, the company gets to keep all the fruits of contestants’ labors; if it had merely sponsored an open-source project, Netflix would have had to share all the innovations that resulted.
The Netflix Prize model will likely work a lot better in the software business than in other industries that depend on intellectual property. That’s because programmers are used to collaborating with one another, even when they work for companies that are competitors. Netflix can be considered a rival to AT&T—both are working on ways to bring movies into people’s homes—but employees of the phone company apparently had no problem helping out the DVD service. You’d be hard-pressed to find such cross-business collaboration in, say, the pharmaceutical industry, where secrecy prevails.
So which tech firms can benefit from setting up prizes? My first candidate is Microsoft. In trying to beat Google’s search engine, the company faces a clear hiring disadvantage—the world’s best search engineers want to work for the world’s most committed search company, and that’s not Microsoft. What’s more, search engines have proved impervious to open-source development; Wikipedia’s founder Jimmy Wales tried to take on Google with his wikilike search project in 2007, but the plan foundered and was eventually shut down, mainly for a lack of interest. How to spark that interest? Money. Microsoft could offer $10 million to the first team that figures out a way to improve its search engine by, say, 10 percent. The difficulty here would be in deciding how to measure the “improvement”; one way of doing so would be to discreetly test contestants’ algorithms on a subset of search engine queries and then to analyze whether users respond to the results.
Google itself could also do well with a prize. The company is heavily invested in solving one of the world’s hardest tech problems—machine translation. A prize would be useful here because translation requires a wide range of expertise—software engineering, linguistics, and a whole lot of math—and a high-profile award could get people from different disciplines to team up. Google could award $1 million for each 10 percent improvement in its algorithm for translating large bodies of text across languages. The improvements would be measured in accuracy—the team that writes code that best translates, say, Proust’s oeuvre into English would win big.
What else? When you start thinking of problems that could be solved through competition, it’s hard to stop. How about a prize for improving face recognition in pictures or videos? Or what about one for improving speech recognition, so that we’ll finally get to talk to our computers instead of type? After all, movies are important—but perhaps Netflix has stumbled on to something much more useful than a way to tell whether you’ll love Napoleon Dynamite.