Findory

As I've written before I am a huge fan of collaborative filtering, and think that it has huge potential.

A company that I thought had a really great idea was Findory. Findory ambitiously aspired to apply collaborative filtering to "information", specifically news websites and blogs. Unfortunately, according to TechCrunch, it seems that the site will go gently into that good night.

Information is a really tough domain for collaborative filtering as it is rapidly changing (as new news and blog posts occur) and relatively sparse (there are many stories, many readers and not necessarily much overlap) It seems that they never got the critical mass of users to be able to find good neighbors for everyone. Meanwhile user rated news sites like Digg and Reddit have gotten huge. The problem with sites like those is that they lead to a sort of mob mentality. It seems that the same type of stories hit the front page and you lose any serendipity. I stopped reading Digg months ago because I knew that I'd inevitably come across anything they had that was truly interesting somewhere else.

Findory had some cool ideas: they tried to use implicit ratings (clicking a link constituted a type of "up" vote) and they didn't require registration for personalization. I used the site quite a bit from October to December 2005. Ultimately I stopped using them because I didn't want to read everything on their site, and their recommendations were fairly boring.

I still think that collaborative filtering could be really useful in the news/blog space. Adding personalized recommendations to digg or reddit based on what people with similar tastes had voted up would pretty much cover it. Both those sites have a large enough user base that finding neighbors wouldn't be a problem.

As Findory founder Greg Linden says in his blog post announcing Findory's demise:

Search only helps if you can say what you want. Personalization helps you discover things that you could not have found on your own.