google + friendster = eurekster?

With all the hoopla surrounding social network sites these days, I was a bit surprised to see that this one, perhaps overshadowed by the launch of orkut, seemed to fly under most folks' radar: Eurekster. Eurekster is a personalized search engine that uses the search activity of your social network to help rank your search results... like if Google took your friendsters into account when computing a personalized PageRank. The site got a feature in ZDNet a few weeks ago.

So I signed up a couple fake accounts to see how it works. The graphic design of the site left me feeling bored and uninspired, so fortunately the registration process was simple and quick, with no long list of interests or demographic data to fill out... just a name, e-mail, and password. Like friendster and any other YASNS they then ask you for a list of friend's e-mails to invite into your social network. You then use Eurekster like any search engine, except that search hits previously clicked-through by people in your social network climb up in the ratings, and are marked with the Eurekster 'e' logo to indicate the previous traffic. Additionally, a side bar on the right of the browser presents the most recent and most recurrent search queries and followed links.

Their business model for the moment is to collect money through paid search results provided by Overture. It's an interesting idea, but I don't expect them to do well. First, these kind of sites need a critical mass to be successful, and I don't think the cost of building up yet another social network will give you a reasonable benefit over what you can already get from Google. Besides, do I really want my friends' search results to influence mine? The voyeur in me kind of likes the idea of seeing the "footprints in the sand" of previous searches, but does it make my search results much more relevant? Perhaps if I could segment my friends into different groups, and apply these groups depending on topic, it might add some relevance, but that incurs even more work on the part of the user. There is also the possibility, especially in smaller networks, of spying other's search queries. Your social network can then play the game of "which one of us has been hunting for porn!?", though to be fair, Eurekster includes a "private search" checkbox that hides your searches from your network. Still, what about people following crappy links, both intentionally (search spamming!) and unintentionally, changing your results? Finally, before I get any further carried away, how big does your social network need to be for it to make any real, significant impact on the majority of your searches?

I think using actual user behavior to improve relevance rankings is a fruitful avenue to explore.... but why even bother with the social network? Why not just use the traffic patterns of everyone, or maybe just yourself? The most useful feature may actually be to have a social network consisting of just yourself, and then all the sites you visited before will rise in the rankings, allowing you to revisit them faster. In the case of one global network, with a little analysis or clustering you could even pick out global patterns and groupings which you could use to refine your search upon demand (though I suppose this is akin to adding user interaction data to a search engine like Vivisimo, which already does clustering). In the end, though, how will any of this successfully differentiate a competitor from Google? That's certainly a tough act to beat today, let alone in the future when Google unrolls their own personalization technology, based on the work they acquired from Outride and Kaltix.