December 28, 2006, 12:00 AM

Netflix personalization challenge draws 10,000 entries to date

The $1 million award seeks improvements to Netflix’s collaborative filtering technology to drive more accurate movie recommendations out of its engine. So far, some 20 entries may represent improvements, CEO Reed Hastings tells Internet Retailer.

In October Netflix said it would award $1 million to the first person or team that can develop a new personalization engine that will improve the accuracy of its movie recommendations by at least 10%. The early response to that challenge suggests that anyone with an eye on that prize has lots of company. So far, CEO Reed Hastings tells Internet Retailer, the competition has attracted entries from more than 10,000 teams and individuals.

“Already more than 20 of them are better than Netflix,” says Hastings of proposals from contest entrants to drive more accurate movie recommendations from Netflix’s collaborative filtering engine. “And they keep getting better. The great news is that we will incorporate all of their improvements into our service so everyone will get the benefit of their hard work."

Hastings adds the company has worked on this project for years before deciding to open it up to other researchers with the prize. “We’re using the tastes of the community to help individuals find movies they will love,” he says.

If no one wins the grand prize in 2006, Netflix will award a $50,000 progress prize for the most significant advancement each year until someone grabs the grand prize.

Netflix is No. 21 in the Internet Retailer Top 500 Guide to Retail Web Sites.

comments powered by Disqus

Advertisement

Advertisement

Advertisement

From IR Blogs

FPO

Jim Erickson / E-Commerce

Why Western brands should register trademarks in China—now

China doesn’t recognize trademarks registered in other countries. Companies that register first in China get ...

FPO

Asher Elran / E-Commerce

E-commerce and duplicate content: solutions to common problems

Google penalizes retail web sites that display the same content on multiple pages, or that ...

Advertisement