Alex Iskold just posted Rethinking Recommendation Engines, a product type that we here at ReadWriteWeb have explored a lot over the past year or so. In this follow-up post, we present 10 recommendation engines that we like. And we don't include the obvious ones, such as Amazon, Netflix, last.fm, Pandora. So it's not a 'top 10', don't panic. We invite you to add your favorites in the comments.
Recommendation engines were included in our toolkit for 2008: What's Next on the Web. Marshall Kirkpatrick wrote that "the future is likely to be even more swamped in data, social and content options than the web is today. From Google Reader's recent incorporation of both feed recommendations and shared items in Reader from your contacts in GMail to the ascendancy of services like Last.fm, Pandora and StumbleUpon - recommendation is beginning to make a big splash already."
Without further ado, here are our 10 picks, compiled from previous ReadWriteWeb posts:
How do you navigate a nearly infinite world of digital data to find the best content for your tastes and needs? Our collective answer to this question is in its infancy, but Oregon based recommendation service MyStrands has raised a whopping $55 million to build on the existing science of recommendation. Definitely the dark horse of the recommendation engines - one to watch.
MatchMine, a Massachusetts company building a cross-platform media recommendation engine, received a $10 million investment from The Kraft Group. The company released an early product called MyMovieMatch in July. See RIA expert Ryan Stewart's review of the original product for background from this summer.
Zync, a Massachusetts-based startup, operates a local event recommendation engine based around the city of Boston. The site currently lists 30,000 events across 20,000 venues. And even though it only has 355 users, they have amassed almost 9500 ratings.
According to Zync, their recommendation technology uses patent-pending algorithms to recommend events, activities, and restaurants to users based on the input of other, like-minded people. Theoretically, as with any peer recommendation system, this one would get better and more accurate the more people use it.
The team behind SeeqPod, a music search and recommendation engine, believes strongly in what they call "playable search." SeeqPod trawls the web, indexing all the music files it finds, and then offers them for playback direct from that location. The company knows that because they are not hosting any music files, but are merely offering links to them, they can neatly sidestep copyright and legal concerns.
Scouta is a web app that provides you with media recommendations, based on preferences and interests you display by your selections within the application. If that sounds complicated, think Pandora, but for all media on the web (including media available outside the US). Or think Last.FM without the fuss about neighbors. To be honest, neither of those comparisons is quite right either. It's more like YouTube, except all the side column content is actually interesting to you.
Music recommendation and discovery engines are hot stuff but what if you could use some of the same juju to better organize the music you already have in your collection? The newly launched Veenix TuneExplorer for Mac does just that. By looking at qualities the company says include "pitch values, pitch variance, fundamental strengths, and a host of other sonic qualities" - the program acts like Pandora within your music collection.
The Filter, a social music recommendation service backed by rock star Peter Gabriel, has released a new version of their software - featuring an improved user interface, a Facebook app and a partnership with Nokia. The Filter is a "playlist creation suite" for iTunes, iPod, iPhone and Apple TV. It works across Windows and Macintosh and it basically allows you to build playlists from the music stored on your PC, Mac, iPod or Nokia mobile phone.
The Filter's user base is reported to be growing at 25,000 a month. The engine can identify 5 million songs, 4.5m of which have clips (short samples). The Filter works by using Bayesian mathematics and it was developed by physicist Martin Hopkins.
Born out of a closet dislike for "Shrek 2," Criticker is a new movie review community and recommendation engine that aims to match users with like-minded individuals who share the same cinematic taste. Once you've rated 10 movies at Criticker it begins to form what they call a Taste Compatibility Index (TCI) that matches you up with not only other users, but also professional reviewers who share your taste in movies (though, we found that site really doesn't start delivering usable results until you've rated around 50 flicks).
FeedEachother is an RSS Reader built by a former developer from Yahoo! Answers and another now at craft social network Etsy. The interface will feel very familiar to anyone who uses Facebook or Google Reader. The service does a good job of communicating for novice users while offering a feature set that power users will really like.
FeedEachOther recommends feeds "similar" to the ones you're subscribed to. Recommendation engines are a key way to leverage the network effect of distributed nodes of knowledge - ala social apps online. Big value there for discovery of high value information sources.
StumbleUpon is a "personalized
content discovery" service, which has grown very popular on the Web. Its main feature is serendipity, finding new webpages by clicking through from other pages 'stumbled' by users. The app is now owned by eBay and it's unknown what they might do with StumbleUpon, but recommending new items to buy might be on the cards.
Now it's your turn: recommend some more recommendation engines in the comments.
Recommendation Resources:
*The Art, Science and Business of Recommendation Engines
*RWW Recommendation Industry Feed Favorites OPML file (save link)
*RWW Recommendation Industry Feeds - Best of Feed (copy and paste to your reader)
*Click to preview the above feeds before subscribing (pop-up window)
*RWW Open Data Sites Search (Visit and Bookmark)
Comments
Subscribe to comments for this post OR Subscribe to comments for all Read/WriteWeb posts
The 'Social Graph Ranking' feature on our service recommends people similar to yourself. You can then check them out on their social networks and find out what media they are into. Give it a try... (Richard please, Richard?)
Posted by: Jack | February 25, 2008 4:43 AMShould have probably added in the comment above that our Loveth.at (http://www.loveth.at) social advertising engine allows users to recommend the brands that they like, to the people most similar to themselves.
Kind of recommending stuff to people recommended as similar to you, if that makes sense.
Posted by: Jack | February 25, 2008 4:49 AMYou missed one but it was easy to miss and thats Amazon.com. I would wager its one of the most used and most sucessful recommendation engines ever built.
Posted by: Darren Stuart | February 25, 2008 5:08 AMRichard,
I agree with Darren. I use Amazon to reseach the customer reviews. I use it more than any other retail product review engine. On websites... I like stumbleupon and delicious. After some use, stumbleupon will really start to get your tastes right. I use delicious as a recommendation tool when I look through other peoples bookmarks and notes. In some ways a bookmark alone is a recommendation. Even if the user hates the site they bookmarked it for a reason.
Posted by: Steve Olson | February 25, 2008 7:22 AMCollaritydelivers automated Amazon-like content recommendations for web publishers. We help them better monetize their content by creating implicit attention communities based on the anonymous behavior of their website audience. This common behavioral relevance foundation is then used to serve content recommendations and advertising.Lots of this information is already sitting in their site log files -- we simply put it to work, in an automated way, for the publisher.
Posted by: Rob Rustad | February 25, 2008 8:25 AMHmm...I guess you've skipped the entire category of implicit recommendation services being applied by merchant Web sites as well as publishing sites like the WashingtonPost.com. @Rob lists Collarity, which I would include in this category as well. Here's a list of the others I'm familiar with in this group:
- Aggregate Knowledge
Posted by: p-air- Loomia (most recently w/SeenThis)
- Baynote
- Certona
- CleverSet (recently acquired)
I suppose I'm probably in the minority here, but as someone who enjoys an eclectic mix of media, the recommendation engines aren't very useful for me. My music collection, for example, ranges from Missy Elliot to Black Flag to Latvian "modern classical" composers. If all these engines can do is recommend music that "sounds like" what I'm already listening to, then they are missing the point: I'd like to hear music that *doesn't* sound like what I'm already listening to.
And for musicians, this can be troublesome as well. For decades, broadcast radio herded music into neatly-defined categories, which musicians needed to adhere to if they hoped to make a living. The decline in broadcast radio listenership and the rise of more free-form internet-based radio gave musicians a glimmer of hope that a more creatively tolerant marketplace was opening up. However, the recommendation engines might be herding them back into the same old marketing categories again.
Posted by: Marcello | February 25, 2008 9:30 AMThis is an impressive and valuable list.
With all of these engines powered by sophisticated algorithms its worth recognizing that we are also witnessing the rise, albeit to a lesser extent, of the human powered web. Our vertical search engine, Search Free Apps, only includes applications that have been individually reviewed for quality and offer either a fully functional free service, or a free trial with no credit card or other ongoing obligation required. Search Free Apps (www.SearchFreeApps.com) now includes over 1,150 distinct free services that range across the full spectrum of online apps.
To be included in this vertical service, an app must meet stringent standards that are enforced by live people. So, this human powered search engine, and other verticals like it, are also a form of potentially valuable recommendation services.
Posted by: Bruce Judson | February 25, 2008 10:23 AMHere is one you missed.
http://www.SpinSnap.com
SpinSnap.com is a Discovery Engine. Most people who use SpinSnap.com discover websites they would not typically encounter in their everyday internet travels.
SpinSnap.com is like StumbleUpon but without having to download a toolbar.
Posted by: Jon | February 25, 2008 11:06 AMThe first para explains that they intentionally did not include Amazon in this post.
We actually have a patented recommendation system that no one really knows about. We tried to get it out before the bubble burst but haven't done anything with it in a few years.
It is a ground floor patent that is basically the opposite of collaborative filtering. It correlates the items rather than the people and is incredibly accurate.
If anyone reading this post is interested in acquiring or licensing it, contact arnie [at] cox.net.
Posted by: Arnie | February 25, 2008 12:05 PMI tend to find recommendation engines most valuable when I am shopping (my tastes in movies, music, and even websites tend to be very eclectic and recommendation engines don't seem to pick up on that well, in my experience). For instance, I like Stylefeeder (www.stylefeeder.com) a lot - it is kind of like Stumble Upon in the sense that the more you use, the most accurate it gets. My Shopping Pal (www.myshoppingpal.com) works kind of the same way.
Another shopping site I like, although it is more of a curated search engine, is 3Luxe (www.3luxe.com). It only offers the three top-rated picks in each category (the picks are based on online reviews, print reviews, consumer feedback, and some actual product testing). It is great when you don't have time to do lots of research.
Posted by: Jacqueline | February 25, 2008 3:58 PMOh yeah, and Cookthink (www.cookthink.com) is pretty good if you're looking for new and interesting recipes.
Posted by: Jacqueline | February 25, 2008 4:13 PMI've been using http://www.streakr.com for a while and it's recommendations seem pretty cool, I've picked up some funny stuff. Seems big with people in from Europe, plenty of pretty british chicks on the network!!!!
Posted by: colin strummer | February 25, 2008 4:28 PMDocoloco is a recommendation engine for local places. It allows users to recommend a place for something. The something then becomes a tag with a rating attached to it and over time you build a collection of recommendations that is keyword searchable and can be presented in a list per tag that can also be filtered by geography and ordered by the collective opinion.
www.docoloco.com
Posted by: Johnny | February 25, 2008 10:05 PMZync is a very promising recommendations engine - it has already shown me lots of venues and restaurants that I didn't know about but match my tastes.
Stylefeeder is the one I turn to regularly for online shopping recommendations. Not only does it recommend products I might like based on my past ratings, but it also shows images of products similar to the one I'm looking at. It's almost magical in that way. The feeling of magic, of "how did it know?" is what, to me, separates the good from the great.
Posted by: Rekha | February 26, 2008 7:23 AMThanks for the list of recommendation engines! They don't seem very advanced though. Almost all seem to be pure social recommendation, which is a technology developed in the last millennium and is inherently very limited.
How is it limited? Well, for starters,
- It only works for most popular items. It doesn't work at all for the long tail. See http://www.boingboing.net/2006/01/05/walmart-apes-dvd-lis.html
for a scary example.
- It doesn't work for new items. Would you like to have the new stuff recommended immediately, not only after X hundred users have already clicked on it?
- It doesn't work or works poorly across media types. For example, if I read news it would be nice to get videos and music tracks recommended based on that, without requiring that a hundred users have clicked on both the recommended items AND the news I read, which is pretty unlikely to happen.
At Leiki (www.leiki.com) our approach is based on the most detailed automatic content analysis (with 60k+ interconnected categories) and automatic analysis of each individuals personal interests. This means all of the above work perfectly - no user needs to click on anything for it to be very accurately recommended.
Social filtering is a nice way to enhance the truly personal recommendations, but it is a very limited basis for a recommendation system.
Posted by: Petrus Pennanen | February 27, 2008 6:55 AMInteresting article
Posted by: i-guide | February 29, 2008 4:06 AMhttp://www.i-guide.ro
how aere u doing i will love to haver a chat with u so pls give me ur msn or yahho id so that i add u and then we have a nice chat there
Posted by: james | February 29, 2008 11:44 AMThank God the developers developers developers thing isn't there! That was all a hoax you know?
This whole Monkey-boy conspiracy thing includes yet another FAKE film, I call it the "Developers, Developers". This my friends simply never happened! This film is not as grainy as the others and was obviously created by some other group than the other two films where they show me jumping, screaming, eating dirt, and generally losing my mind. This one had a higher budget, we suspect it had corporate backing (aapl). The actor in this one looks a lot like me, (Peter Boyle called me once and said it was him, he apologized, he said at the time he was being blackmailed so he did it). If that is not good enough, there is one dead give away: THE SWEAT! You see I do not sweat, I was born with a rare condition called Anhidrosis Glandofelimumia, I cannot sweat. I take medication weekly and always stay in cool environments. If I got hot like this guy I wouldn't sweat, I would swell up like a water balloon. The conspirators did not know this little known fact about me.
Posted by: steveballmer | February 29, 2008 10:47 PMSpread the word, get the truth out, that's why I blog!
Library Thing is great for recommendations and purchasing ideas!
Posted by: Jill Egan | March 4, 2008 4:29 AMhttp://www.librarything.com/
Many of these music recommendation engines actually narrow your discovery by slowly focusing on what you like.
For people who like a wide variety of music there's a product from Dash Media that filters Internet Radio that actually expands as you use it, so your stations never get old.
Posted by: Joe | March 10, 2008 4:27 PM