Facebook’s new Graph Search feature is still in a very limited private beta release, and only a miniscule number of its millions of users have had the chance to try it out. Even so, over the past week, Graph Search has been widely discussed, hyped, praised, and panned. At one end of the spectrum, commentators are suggesting that Graph Search is Facebook’s first serious challenge to Google’s search engine model, while at the other end, observers have called it everything from a rehash of an old feature to a boring non-announcement or yet another somewhat crass effort at monetizing user engagement.
Like most Facebook users, I have not yet been granted access to Graph Search. But I’ve had a chance to play around with the new tool. Just a few minutes after posting an inquiry about it on Facebook, a friend commented that he had access to the beta version and would be happy to show it to me. In this simple action, the potential of Graph Search has been revealed, though the first iteration of the tool falls short of that. Simply put, there’s a huge corpus of data buried in the social network that is of interest for search and discovery, and Facebook has made a serious move toward surfacing that content for users.
As you’ve no doubt heard in press reports since last week’s announcement in Menlo Park, Graph Search allows users to enter natural language queries related to “likes” and interests shared by friends and others on Facebook. (You can sign up for the private beta version here.) In this first release, the use cases are limited by design to “people, photos, places, and interests” so “friends who have access to Graph Search” is pretty far outside its current bounds. But even in that limited set of data points, the potential for structuring interesting queries is pretty broad. You can search, as Mark Zuckerberg did in his keynote address, for photos of yourself with your wife, and Facebook will return an attractively formatted results page with every photo you’ve ever posted that meets that description. Just in terms of surfacing your own content and content related to your friends, Graph Search proves its usefulness quickly and decisively.
It’s interesting to observe what Facebook means by “natural language” search, however. Zuckerberg joked during the keynote that the company’s first idea for Graph Search was a screen filled with dozens of drop-down menus, only to emphasize how much better and more intuitively the natural language interface behaves. And yet Graph Search strongly emphasizes a highly structured syntax, going so far as to auto-suggest queries that will work well within its sharply defined structure. It really isn’t so different from a series of drop-down menus after all. You are encouraged to type things like the following:
- Restaurants my friends like
- Restaurants near me that my friends like
- Restaurants near me that my friends who like Pokemon like
- Restaurants near me that my friends who like Pokemon and work at Starbucks like
And so on. You can even search for things that have no obvious utility for normal users, like “non-friends who like restaurants near me.” As is the case with much of Facebook’s architecture, Graph Search is surprisingly explicit about treating social interconnection as a collection of binary values in a database. In its way, Facebook has created a semantic universe within its social network, with explicitly defined entities like people, places, and interests that can be connected using explicit marks of affinity: friends, likes, followers, check-ins.
What Facebook is calling “search” is really a just a user-friendly way to query that database. Clearly, all of Facebook’s careful messaging about privacy notwithstanding, Graph Search is designed to return a result for any well-structured query as long as the content in that result has not been explicitly tagged as private. My friend was able to search for “photos uploaded by [Random Person Who Is Not My Facebook Friend]” and other similar queries that had a fairly high creep factor.
The highly structured syntax of Graph Search might lend itself quite readily to a voice-based search interface like Siri, though users must first become comfortable with that syntax or the experience could be frustrating. I also see a strong linkage between Graph Search and Facebook Nearby as methods for surfacing local content, so I wouldn’t be surprised to see Graph Search rolled out for mobile in the near term.
Despite what its detractors have said, Graph Search will no doubt become an integral part of the Facebook experience, and anything that makes users uncomfortable will likely be tweaked as the tool is launched. The focus on places as a primary search entity means that Graph Search could have big implications for local. Businesses would be well-advised to get ahead of this trend and make sure they are doing everything possible to create and maintain an engaging presence on the social network.
There will be growing pains. My friend and I were able to turn up many results for restaurants, a few each for dentists and bowling alleys, but none for plumbers. Among the results we did return, many were not close by or else there was no clear indication why these results and not others were shown. As others have suggested, broad usage may only occur if users are engaged by the search experience and thereby encouraged to share more content about local businesses in the form of likes, check-ins, and ratings.
Partly, too, these results demonstrate how deeply we’ve been trained by organic search following the Google model. Given that Facebook is using multiple individualized cues to return search results customized for users, one clear observation is that traditional SEO, already threatened by Google’s moves toward personalization, will be close to meaningless in this new context.
Graph Search has broad implications for Facebook and the search industry in general, only some of which I’ve touched on here. I’m sure I’ll be returning to this topic in upcoming posts.
Damian Rollison is vice president of product and technology at Universal Business Listing, a company dedicated to promoting online visibility for local businesses. He holds degrees from University of California, Berkeley and the University of Virginia, where he worked at the Institute for Advanced Technology in the Humanities. He can be reached on Twitter.