By now we’re used to music discovery and recommendation engines, but what if I told you that there’s a site that learns your tastes and makes recommendations on everything: music, movies, books, websites, restaurants, and everything else.
Two interesting things are: (1) that the items in the Hunch database are submitted by the users themselves, and (2) what you like is calculated within a single overarching algorithm, meaning that the movies that you like will reflect on your recommended books or music or anything else.
Hunch also has a social media component where you can find ‘tastemates’ whose taste are similar to yours, you can follow people and ‘influencers’ and add your Facebook friends, etc.
Does it work?: it seems to. I would rate the quality of recommendations served to me as generally good. Of course, there’s always the caveat that with more ratings the recommendations would be of better quality, and that the recommendations may be a lot more impressive a few weeks down the road.
The user experience: is quite pleasant. You can choose whatever category you want (movies, music, fashion, restaurants etc.) and be served suggestions. Mouse over any thumbnail and you will see the rating they predict for you (on a 5-star scale). You can ‘educate’ the algorithm by entering an actual rating. The more you do this, the more accurate the recommendations (presumably). The screenshot below shows prediction (left) and user rating (right).
The one thing I noticed is that if the star prediction is right, I as a user am unlikely to re-rate it and confirm, depriving the algorithm of valuable input. They should have a ‘confirm rating’ link or something similar. I also wished they would allow half-stars.
Viewing recommendations right on the page: a lot of music/videos can be played right in on the website itself, via a floating player, which is very cool, especially for publicly available videos. You can also play trailers for recommended movies on the spot. If you like anything you can save it for later viewing at the click of a button.
‘Community’: you can have Facebook or blog style commentary/discussion on everything (see screenshot below).
Moreover, you can find users with similar tastes and follow them, or browse their recommendations, etc.
The verdict:
‘Hunch’ is intriguing, and rather enjoyable to use. Recommendation engines work when you take the time to feed them with information, and actually rate the things they serve. I like the idea that I can invest my time and energy in a single resource that can make inferences about what books I might like by drawing on the data I entered for movies or other categories. I also like the fact that the items presented are recommended by users.
Hunch is surprisingly well put together. I am surprised that they do not have mobile apps though, which seems like a must these days, especially for an outfit that was acquired by eBay and therefore presumably has the resources. Try it out and tell us if you like it.
