Matchmine, of Needham, gathers data on users' preferences to let websites better customize their offerings.
Matchmine fine-tunes its approach
It's easier to track online favorites
Matchmine, of Needham, gathers data on users' preferences to let websites better customize their offerings.
A company that tracks Internet users' preferences recently landed five new website partners, tweaked its technology, and altered its strategy to entice more people to use its service.
Matchmine uses algorithms to gather data on users' likes and dislikes to allow sites to better customize their offerings. The hope is that the more accurate a site's recommendations, the more traffic it will see. So far, a few thousand people are using MatchKeys.
Users create MatchKeys to record their preferences, based on ratings they give music, movies, online videos, and blogs on partner websites. The key uses the information to map their tastes and unearth other content users are likely to enjoy.
Matchmine, of Needham, went live in March with two websites, Fuzz.com, a music site, and FilmCrave.com, a movie site. Five other sites have signed on and should be up and running this quarter.
The company has made it easier for users to create MatchKeys, adding an icon to its partner websites that people can click on to create keys, rather than going to a separate site. People can store their MatchKeys online, increasing their portability and eliminating the need to download software.
"We had a lot of trouble driving people to matchmine," said chief executive Mike Troiano. "We shifted gears, said 'The right way to roll this out is to get people to something they want more quickly.' It's really that change in strategy which led to our first five deals."
Matchmine's new partners are:
Although there are other discovery engines on the Web, matchmine is different because it records someone's likes and dislikes across different sites and different types of media, said Jeff Yasuda, Fuzz.com's founder and chief executive.
Because the MatchKey casts a wide net, the recommendations it makes are more accurate, he said.
"There's a secret black magic involved with how exactly discovery takes place," he said. "The feedback we've heard is that it's pretty cool and highly useful. The recommendations it's made have been great."
The method matchmine uses has more detail about what an item is than first-generation engines that mine content for users, Troiano said. Those engines use a system called collaborative filtering. It determines most people who like songs A and B also like song C, but the system doesn't know exactly what the songs are; matchmine acquires all the information about an item, such as title, any comments attached to it, or whether anyone has tagged it.
"The starting point of our system is really understanding at the individual item level what are the attributes," Troiano said. "A system that is flexible, a system that acquires as much data as possible; those are the two cornerstones of our strategy."
Matchmine's approach helps eliminate what's known in the industry as a cold start. Sites have a limited amount of information on users at their first visit, said Joshua Kinberg, Odeo's product development vice president. The site has data only about what the user has purchased, listened to, viewed, or download from that site, but not others.
Because the MatchKey is portable, Odeo gains more information about its visitors' tastes and can offer more accurate recommendations, Kinberg said.
"You just have more user behavior to create more detailed rankings," he said. "If a user is able to discover cool stuff, that user's going to return. That's how you build a loyal user base."
© Copyright 2008 Globe Newspaper Company.
