Video clerks power new movie recommendation engine

Even as technology threatens the survival of video rental stores, serial entrepreneur Stuart Skorman thinks there's still a place for the movie-matchmaking advice of veteran video store clerks.

To prove his point, Skorman hired more than 20 former video store clerks to pour their collective wisdom into a new Internet search engine, called ClerkDogs, that's being unleashed Tuesday.

Skorman is betting this combination of human intelligence and data crunching will emerge as a more engaging and intuitive alternative to the highly automated movie-recommendation system that has helped fuel the success of online DVD rental leader Netflix Inc.

Netflix spits out recommendations for its 8.7 million subscribers by drawing upon a database of 2 billion ratings that its customers have entered during the past decade. By knowing whether you liked one movie, Netflix suggests others you might enjoy by mining the past preferences and renting patterns of subscribers who watched the same movies.

ClerkDogs also dives into a vast data pool in order to suggest movies, but augments its findings in key ways.

After it asks visitors to enter the name of a movie they liked, ClerkDogs' engine generates a list of suggestions based on a computer-driven analysis of video clerks' insights and written reviews. And for a more personal touch, it lets its users tweak recommendations based on their moods at the time of a request. ClerkDogs users can slide a scale indicating whether they are looking for movies with a little more romance, suspense, humor and other elements contained in the movie they initially selected.

The revisions help ClerkDogs get even closer to a visitor's interests, pulling from the information provided by the video store clerks.

Skorman, a former video store owner, likens the system to having a conversation with a movie buff — something that's not possible on Netflix's recommendation engine because it relies on a strict one- to five-star rating system and doesn't adjust for a renter's changing emotions.

"The reasons people choose certain movies are such a complex thing," Skorman said. "Netflix has done a wonderful job with its system, but it has gone as far as it can go."

Netflix's recommendation engine accounts for about 60% of the service's rental requests, a sign that it's doing a pretty good job divining what customers like, Netflix spokesman Steve Swasey said.

Even so, in hopes of making its recommendation engine even more effective, Netflix has been offering a $1 million prize to anyone who can develop a software program that improves upon its matching system by at least 10%. After more than two years, the leading entrant in Netflix's contest is within 0.44 percentage points of hitting the target.

"That final stretch is going to be like going up the final 300 yards of Mount Everest," Swasey said.

This isn't the first time that Skorman, 60, has tried to build a better mousetrap for movie lovers.

In the 1980s, he took over a video store in Vermont and expanded it into a successful chain that he sold to Blockbuster Inc. for $3 million during the early 1990s.

A few years later, he moved to the San Francisco area and started Reel.com, an online service specializing in selling movies on videotape and, of course, recommending movies. Hollywood Video bought Reel for $100 million in 1998, generating a $17 million windfall for Skorman.

Page
  • 1
  • |
  • 2
Join the Discussion
blog comments powered by Disqus
 
You Might Also Like...