Wants recommendations on how to improve.
Online DVD rental pioneer Netflix Inc. wants recommendations on how to improve its movie-recommendation system so badly that it's dangling a $1 million reward as an incentive.
The prize, offered in a contest scheduled to begin Monday (Oct. 2), is part of Netflix's effort to sharpen its competitive edge as it continues a bitter duel with Blockbuster Inc. and prepares for an anticipated onslaught of services that make it easier to download movies on to computer hard drives.
By spurring engineers to develop a better way to decipher consumer tastes, Netflix is betting its market-leading DVD service will become even more useful to its 5.2 million subscribers and attract new customers.
The concept mirrors a similar contest run by the federal government's Defense Advanced Research Projects Agency to promote the development of automated vehicles that could be used on a battlefield. A Stanford University team last year won a $2 million prize from DARPA by developing a robotic car that covered 132 miles of desert terrain in just under seven hours.
To win Netflix's prize, a software program must improve upon the DVD service's current movie-recommendation system by at least 10%. The differences will be tracked by a program that quantifies how well the recommendation systems predict which movies will be liked or disliked by a profiled consumer.
"Right now, we're driving the Model T version of what is possible," said Netflix Chairman Reed Hastings. "We want to build a Ferrari and establishing the Netflix Prize is a first step."
The winner will be free to license the program to other Web sites, such as Amazon.com Inc., that make product recommendations to their visitors.
Recommendation software is expected to play an increasingly important role in electronic commerce as Internet companies expand their databases of past consumer behavior. Netflix already has learned a lot about DVD tastes, having gathered 1.5 billion movie ratings from about 10 million consumers who have subscribed to its service at some point in its seven-year history.
As they stockpile more information about each individual's habits and preferences, online businesses like Netflix, Amazon.com and Google Inc. theoretically will be able to make more money by anticipating what consumers want before they even realize it.
To fuel its contest, Netflix is releasing 100 million movie ratings pulled from about 500,000 of Netflix's current and past subscribers, said Jim Bennett, the company's vice president of recommendation systems. He said all personal information has been removed from the data so there is no way participating engineers will be able to identify the names associated with the ratings.
The data set is so large that it will take about 20 minutes to download on a high-speed Internet connection.
Netflix doesn't think improving on its current system will be easy so the company will pay $50,000 annually to the contest entrant that's making the most progress toward the final goal.
Entertainment industry analyst Dennis McAlpine doubts Netflix is going to get much bang for its buck.
"I'm a little surprised because their current system already works so well and the (customer) satisfaction level is pretty high," he said. "This looks like a marketing gimmick."
Copyright 2006 Associated Press. All rights reserved. This material may not be published, broadcast, rewritten or redistributed.