Tuesday, September 22, 2009

Netflix Launches Second $1 Million Search Contest

Just after announcing the winner of its $1 million Netflix Prize for improving Netflix search, Netflix announced a second challenge. The first Netflix Prize went to a team that improved the Netflix movie-recommendation system. Netflix Prize 2 focuses on a tough challenge: Predicting movie enjoyment by members who don't rate movies often.

Netflix on Monday announced the winner of its $1 million search contest. Just moments later, Netflix launched a new million-dollar challenge to encourage engineers, computer scientists, and machine-learning communities to keep working on improvements.

After three years and submissions by more than 40,000 teams from 186 countries, Netflix awarded the $1 million prize to the team that most improved the Netflix movie-recommendation system. Specifically, the teams set out to improve upon the company's ability to accurately predict Netflix members' movie tastes by 10 percent -- a hurdle Netflix scientists were not able to overcome on their own over the last decade.

Netflix cofounder and CEO Reed Hastings said it was a bona fide race to the end, with teams that had previously battled it out independently joining forces to surpass the 10 percent barrier. "New submissions arrived fast and furious in the closing hours," Hastings said, "and the competition had more twists and turns than The Crying Game, The Usual Suspects, and all the Bourne movies wrapped into one."

Improving Netflix

When Netflix launched the Netflix Prize in October 2006, it made 100 million anonymous movie ratings -- ranging from one star to five stars -- available to contestants. All personal information that could identify individual Netflix members was removed from the prize data. The data contained movie titles, star ratings, and dates, but no text reviews.

Accurately predicting the movies Netflix members will love is a key component of the Netflix service. Neil Hunt, Netflix chief product officer, said this extreme level of personalization is "like entering a video store with 100,000 titles and having those that are most interesting to you fly off the shelves and line up in front of you."

Netflix Prize 2 focuses on a much tougher problem: Predicting movie enjoyment by members who don't rate movies often, or at all, by taking advantage of demographic and behavioral data carrying signals about the individuals' taste profiles.

Netflix's Five-Star Move

Unlike the first challenge, the new contest has no specific accuracy target. That's because Netflix and contest judges have little idea how far experts can push the data to drive useful predictions. For this reason, $500,000 will be awarded to the team judged to be leading after six months, and an additional $500,000 will be given to the team in the lead at the 18-month mark, when the contest is wrapped up.

Greg Sterling, principal analyst at Sterling Market Intelligence, called the Netflix contest a great move all the way around.

"Crowdsourcing the Netflix algorithm, getting a better user experience as a result, and all the positive PR from the contest. It's an example of Web 2.0 best practices, although that term is now passé," Sterling said. "One million dollars is nothing to Netflix, and it's a big enough prize to get some top-notch folks involved. It could potentially even result in some engineering hires down the line."