Local taste-grapher Hunch recently rolled out a feed of social recommendations and made it the default landing page, which has confused some users who wonder how it differs from sites like StumbleUpon and Digg, which also rely on the social graph and user feedback to serve recommendations. The crucial difference? StumbleUpon and its ilk emphasize web content–news, websites, apps, videos, especially whatever memey thing is trending that day or week–while Hunch is equally interested in recommending restaurants, events and books.
Hunch also thinks they’re able to start making spot-on recommendations right away by requiring Twitter, Facebook, Foursquare or LinkedIn authentication; if the user doesn’t have those things, Hunch asks them ten questions and starts building a taste profile based on those. StumbleUpon is hip to social authentication as well, but users can still sign up with just an email address.
The questions–“Tell Hunch About You”–are another notable difference between the sites. Rather than asking you to select verticals you’re interested in, Hunch asks you to play a little game. The questions are so quick and easy that I had apparently already answered 218, which I don’t really remember doing. I also just answered two more. The average user has answered 113 questions, according to Hunch.
Hunch thinks their strategy gives them a better chance of predicting taste for anyone, including people outside the majority demographic of Hunch users (male, college-educated, white, is what we’re guessing it looks like at this point).
StumbleUpon may have many more users and thus a more robust database of social data and user feedback at this point, but Hunch’s technical foundation is more hardcore. It takes Hunch two days to fully refresh its taste graph, despite running it on a 48-processor supercomputer with a terabyte of memory. Further: Hunch’s taste graph infographic, posted last week, is way more complicated than StumbleUpon’s corresponding “recommendation technology” infographic.