Facebook Graph Search is super-powerful – if all your friends obsessively post to Facebook

Facebook’s new Graph Search is an exceptionally powerful idea. Here are some searches I’m looking forward to running:

• Bars my friends like in San Francisco
• My friends who like Doctor Who and live near me and like pizza
• Friends that know [insert investor or entrepreneur name here]

Make no mistake: this is a new kind of search that shows the way for more advanced social software, everywhere. There’s no doubting that it’s an impressive achievement, that has the potential to change the way people use the Internet.

Or it would. You see, Graph Search’s Achilles’ heel is Facebook itself.

Facebook is a walled garden; a closed box. Information on the service stays on the service, and it’s hard to import it in from other places. As a result, the information about your likes, your friendships and your location in Graph Search are needfully based on information you and your friends have explicitly posted to Facebook.

That makes it much less of an achievement than it could have been. To use the above examples, I don’t tend to like bars or restaurants on Facebook (although I might on Yelp or Foursquare, or mention them on Twitter); I’m not going to tell Facebook that I like pizza, because why would I; and I’m much more likely to “friend” an investor or an entrepreneur on LinkedIn than Facebook, because the former focuses on my work achievements, and the latter is exponentially more likely to show them a picture of me at some random party next to some guy I don’t know wearing one of those beer hats. (Reader, I speak from experience.)

So Facebook is working from an incomplete social graph. What’s interesting to me is that if they turn more of their focus to social graph search, it makes more sense for them to start capturing more of the outside world – and become more open in the process. A company that to date has spent most of its energy capturing esoteric information about peoples’ personal lives and locking them in a black box will now have to learn to talk to the rest of the web, in order to allow this product to reach its full potential. There’s even an opportunity here for crawling and encouraging federated social networks.

Will this happen? It’s hard to say. Whether they decide to open up and start consuming and publishing graph information depends on their internal culture. It’s also a more complicated problem than, say, implementing a Google-style page search: Facebook has access permissions, which must be obeyed. The search results I see might be very different to the search results you see, based not just on our different social graphs, but what information about themselves our friends have chosen to allow each of us to view. If Graph Search just crawls public information, this is moot, of course, but having this deep level of privacy integration would be an actual reason to use Facebook to store this information. (Or for Facebook to start exporting its access control to third parties via an open API – this was an obvious route for Google+ to take, but they’ve been surprisingly slow to do so.)

Until Facebook opens up, which given all available information is a bit like saying until hell freezes over and golden rain falls from the sky, there’s still everything to play for in social search. Facebook has shown us the way, but while they only use a tiny subset of the information available to them, it’s only a proof of concept.