We last covered Matchmine in September 2007, noting that the product, a preferences and recommendation engine, was on to something very interesting. They were just going about it the wrong way. I'm glad to report that company has seen the error of its ways and, while keeping true to its mission, now has a product that makes sense.
To recap: Matchmine will tell you what media you will like (it covers blogs, music, video, and movies) based on what you tell it you already like. It's the same idea that you see every day in Netflix, except the Matchmine technology is more robust and accurate, according to CEO Mike Troiano. And, more importantly, the preferences data are portable.
Inside the Matchmine database: The key to getting you to rent more movies is knowing your specific tastes.
Here's how it works--or at least how it should work once Troiano gets the business fully up to speed: You tell your media site what you like, just as you do right now, and then the super-duper Matchmine algorithm calculates the "distance" between your items and other items in its data model to find new things for you that you'll like. Troiano says that even a small improvement in the preferences engine can have a dramatic impact on sales at a media site, and that Netflix, even with a prize offered for the solution, has not been able to improve its engine 10 percent in one year of trying. Matchmine does better, he says.
Individuals' Matchmine "keys" are stored on Matchmine's servers, not in a local file that requires an executable to access, as in the last, wacky version. This is good. It means that sites that want to participate in the Matchmine system simply have to strike a deal with the company to be able to offer up the cross-site recommender to their users. They don't have to get the users themselves to install anything.
It will be hard to get market-leading commerce sites (Amazon, Netflix, Last.fm, etc.) to adopt this product, since they no doubt see their users' preferences as strategic assets. However, if Matchmine can actually improve their recommendation success metrics, it may become more strategic for them to adopt the technology than to shun it.
The company is also hoping that consumers will continue to demand personal data portability. People don't want their preference data locked into commerce sites, Troiano maintains, even if the sites don't want to let it go. This system might just satisfy both sides of the market: It would give users a way to take their likes and dislikes with them, while at the same time making them less inclined to do so, since the sites that use Matchmine would offer a better experience because of their deeper understanding of customers' preferences.
Because of the strategic change Matchmine needs to effect in its potential customers, it will be a difficult company to grow. But difficult is good--it means it won't have dozens of me-too companies hustling for the same business.
You can try Matchmine on the music site Fuzz and the film site FilmCrave.
We gave Matchmine a small writeup from the DemoFall conference yesterday, but I wanted to dive into the concept here a little more. What Matchmine is trying to do is create a universal preferences system. The pitch is that instead of rating the movies and music and blogs you like on each site you go to that has ratings, you rate your preferences once, and then any future sites you go to can grab those prefs immediately to serve you recommendations that will be good for you.
By telling Matchmine what you like, you create a key that can be used by other sites.
The recommendations are created independently of any site's users, and thus can't be based on collaborative filtering, where you're matched with items based on what other users are buying or rating. Instead, Matchmine finds items for you based on a database of attributes attached to each property in its database. If you rate Western movies highly, it will find more of them for you. If you're partial to Oscar winners, it knows that, too. Of course, the more items you rate, the better your results should be.
The company plans to make money be selling its technology to content marketplaces, and it already has deals with some second-tier music and film sites like Fuzz, FilmCrave, and Peerflix.
There's a standalone recommendation application that connects you to Amazon and other stores to buy things.
These sites, though, already have their own databases of media as well as their own rating systems. Sites that support Matchmine will let their users import their preferences, but they don't, yet, export. So any work you've done on Peerflix, for example, to rate your favorite movies won't get exported to your Matchmine preferences set. And it's unlikely that the major commerce/rating sites, like Amazon and Netflix, will adopt technology that commoditizes what for them is a key service. I really don't think Amazon sees its recommendation engine as something it can outsource.
On the other hand, Matchmine can send transactions to any store it wants, including Amazon and Netflix, and pocket affiliate fees for doing so. The service's standalone recommendation engine does just that.
One of the things I don't get about Matchmine is that it stores all its preference data locally, on users' machines, in a background executable file that the user then authorizes applications to access. This lets users control their information, but it's a very un-Web concept: It's not portable and it sucks up resources on the PC. The service also makes a big deal about privacy, since it doesn't store any personal information with your preferences. But as soon as you import prefs into a commerce site, you'll be matched up. So I don't quite get that.
Matchmine is a noble experiment. The company is trying to take very personal information about users--their likes and dislikes--and give that data back to us for us to use and control as we wish. That's fantastic. It looks like it's going to be difficult to make it work as a business, though.
- prev
- 1
- next
