Facebook has introduced a new kind of search tool for its unique trove of data. It's less useful than it appears to advertisers, users, or developers.
By now you've heard that Facebook announced a beta of a search tool they are calling Graph Search. Never mind that Wall Street was waiting less than patiently for news of how Facebook plans to monetize mobile; Graph Search ignores mobile, being built for the desktop only. (By the way, one of the architects of Graph Search is Lars Rasmussen, who defected from Google in 2011 after his project there, Google Wave, was killed. He also pioneered Google Maps.)
The initial rollout is happening slowly; Facebook is calling this a beta, after all. This search will do less than you might suppose at first; but it will also do things no search has done before. Graph Search lets users (and others) explore the connections among people, places, photos, and interests. It does not search the text of users' Facebook posts. The "signals" (to use a word familiar to fans of Google's search mechanics) that Facebook's search construes are few in number: things like tags, Likes, and relationships.
Here's how the mechanics of the natural-language front end work, according to an Intel Developer Zone blog:
Users can search across [people, places, photos, and interests] using verbs that also work as search modifiers: lives, likes, works; nouns, such as friends, restaurants, New York, pizza; prepositions such as before, with, or in; pronouns, such as who, him, or her.
That privacy thing
There is disagreement in the accounts of Graph Search I have read as to whether users have any effective privacy controls to limit the searchability of their content. A Forbes blog claims that Facebook has unwound the ability of users to declare their profiles not searchable. The Intel blog linked above, and other sources, indicate that users can mark content private and it will not be indexed for search. Given Facebook's history of constantly ratcheting up the amount of user data forced into the open, I know which version I'm inclined to believe until greater clarity emerges.
The "con" of Graph Search
Steve Cheney's blog delves into Graph Search from the point of view of how useful its "signals" are. Primary among them is the Facebook Like. Cheney, coming from a background in the advertising and marketing spheres, knows full well how may Likes have been outright purchased by brands over the last 4 or 5 years. The network of Likes represents unreliable ("dirty") data, and search that relies on that data as a signal is born compromised.
Advertisers that rely on Graph Search to target ads will get results that are no better than the underlying connectivity data -- that is, no better than the validity, and up-to-dateness, of Likes.
What's in it for developers?
Facebook hasn't released any details of an API that would give developers, and their Facebook apps, access to Graph Search. The devil will be in the details as to how much of the tool's power Facebook is going to expose.
The main benefit of Graph Search for developers will be its effect on application discoverability. This applies not only to writers of Facebook apps; in fact any app's Facebook page (or indeed that of any product whatsoever) will end up being more easily discoverable. This is because Graph Search gives users more ways to find out what their friends Like.
Therefore, in order to get the maximum benefit out of the new search, developers are going to have to make sure that their fans declare their loyalty in the form of Likes. And they would be well advised to groom the quality of these Likes -- to avoid, for example, buying Likes by giving away Starbucks cards. However, the quality of the results that searchers get from Graph Search is going to be limited by the inherent unreliability of the connectivity of the underlying graph.