1 Comment

Summary:

Hundreds of thousands of users already use Facebook’s Graph Search tool, a product manager said at a briefing for reporters Thursday. But a lot of challenges are ahead as the company develops the product further.

Screen shot of slide from Facebook presentation on Graph Search tool
photo: Facebook

Facebook’s Graph Search, a tool released in beta last month, sounds like it could permit users to tap the intelligence of friends, and friends of friends, and so on. Once on it, you should be able to get tips on restaurants, music and movies from people you trust, or people who went to certain schools, or lived in certain countries.

But what if the product is not quite ready? In some cases, search results might simply be disappointingly limited — for example, people who said on Facebook that they are from Japan “like” just two restaurants in New York City. That’s what Facebook engineer Mike Curtiss was able to pull up during a Graph Search demonstration for reporters on Thursday at the Facebook campus in Menlo Park, Calif.

Search results pop up on your screen within a couple of seconds, thanks to the Unicorn search engine Facebook developed. The system’s method of connecting drilling down on node after node of the users and cultural items that matching queries appears to be a smart way of refining queries again and again, and Facebook’s flash-only database servers can support that activity in very short order.

But it’s not so much about the technology behind Graph Search: Limited search results are just one example of the tool’s current shortcomings for users. Even Facebook founder and CEO Mark Zuckerberg has asked the company’s search-infrastructure team to improve the Graph Search, Curtiss said.

So far, hundreds of thousands of people are using it, said Tom Stocky, a Facebook product manager. “It’s a long way to go between that and a billion,” Stocky said.

Graph Search only works in English, and even that presents plenty of challenges. Since it debuted last month, Facebook has started supporting ways of calling up friends when people use words other than “friend” in the Graph Search box — homies, besties and the rest. But the tool doesn’t distinguish between people who live in a given city and those who lived there in the past. And it doesn’t filter out people who are near certain states. Likes and check-ins provide data for Graph Search to work with; status updates are not yet searchable.

What’s more, plenty of Facebook users have taken certain information off Facebook or don’t update it anymore with real work information or their favorite movies. In other words, the quality of the data feeding the search results holds back what Graph Search can do.

Facebook first started working on Graph Search two years ago. It’s not like every Facebook engineer has gone to work on the function, but some resources have been allocated to it, and that clearly will continue at least in the next few months.

And there’s a lot to do in the way of query optimization, or structuring searches in ways that will yield better results, Curtiss said.

The big question is whether the efforts going into Graph Search will pay off for Facebook in terms of bringing excitement among users back to the site, or whether Facebook will try to monetize Graph Search by giving businesses access to the tool. That could make Graph Search pay off in a big way.

You’re subscribed! If you like, you can update your settings

  1. I think your friends helping you discover new things is the future of search
    but just discovering from random friends isn’t really effective

    There is a facebook recommendation engine app that is trying to help filter friends to improve discovery called recomendo.
    http://recomen.do/

    it recommends facebook likes for a user based on their taste and in addition.
    you can designate friends as tastemakers in specific categories and that influences what is recommended to you.

Comments have been disabled for this post