Why I could, sort of, Like Graph Search / by Dan Weingrod

I’ve become a very reluctant user of Facebook over the past couple of years. I log in once a week at best, ignore the weekly updates and never sign in to anything with Facebook. At this point I’m down to three, pretty lame, use cases for Facebook:

  • Spying on my, adult, children (pathetic)
  • Following political news and posting views to a broader network than on twitter (this kind of ended with the election)
  • Using it as a version of Patch.com to find out what’s going on locally because I don’t follow local friends on twitter, (like when we lost power in the freak Halloween storm over a year ago).

So when Graph Search launched I pretty much tried to ignore the Apple-like, shrouded in secrecy, intro event. But as I began to think and read more about Graph Search I realized that there’s potentially more to like, than there is to dislike.

For starters there’s the fact that Graph Search could be, as Danny Sullivan pointed out, a fundamentally different kind of search. It’s not the Google type search we may have been looking for or maybe expected and that’s kind of exciting. Sullivan calls it “multidimensional search” and John Battelle thinks of it as “Facebook is no longer flat”. The dimensional metaphors make a lot of sense. When you consider the possibilities of Graph Search you can see that it has the potential to add additional, and potentially very interesting, layers to the Facebook experience. And unlike recent Facebook copycat clones, (I’m looking at you Poke), there’s some serious thinking and innovation going on in terms of deployment of natural language search and linking volumes of structured and unstructured data on a massive scale.

But beyond the potential dimensionality of the Facebook experience there’s also the fact that Graph Search feels like a serious attempt to build a serious model for sematic search. We’ve been talking about sematic search for quite a while, and while there have some halting attempts, this feels like the first time someone is really trying to approach this in a committed fashion. So thinking about Graph Search as some sort of awesome Big Data project it actually begins to feel interesting. Perhaps by  drawing connections and inferences from all of these data points we can learn how people connect and maybe make all the Facebook experience a bit more interesting? Maybe Graph Search could be an alternative to what has become quickly a very tiresome stream.

Of course the real question is, could that even happen? As Steve Cheney pointed out: “much of the structured data that makes up Graph Search is…:totally irrelevant and dirty.” With all the years and dollars spent on buying “Likes”, a great deal of the semantic data in the Facebook ecosystem is pretty much polluted. It’s as if Google had launched organic search AFTER having deployed paid search, and then used paid search data as a basis for ranking.

All of this brings up the issue of the use cases for Graph Search. We’ve seen few great examples of “Stupid Graph Search” tricks like: Mothers of Jews who like Bacon on this Tumblr. And We’ll keep seeing tricks like these for a while to come reminding us of the pitfalls of semantic search within the Facebook environment. Between paid Likes and the “innovation” of frictionless sharing there is going to be a need to focus more effectively on privacy and the inadvertent settings that have become as part of the Facebook experience. And this can’t simply be the role of Facebook users, Facebook itself and the Graph Search team may have to play a bigger role in deciding how deep trolls and how relevant it makes the connections. The idea of creating “obscurity” on Facebook, as discussed in this recent article, may also be a role that that Graph Search will need to take on, on behalf of the users. (And maybe the impetus to do that would be to start thinking of them as customers instead of, how I just wrote, users). By deciding how much and what type of data to relate or interweave Facebook itself can help create a meaningful obscurity. This is a very tough problem, but it’s the responsibility Facebook has accepted by creating Graph Search and in a way, it would be pretty exciting to see them solve it.

All of which leads to the question of the use case for Graph Search. When I first heard about Google it was “there’s this search engine that gets it right and does it really, really fast”. Right now I haven’t heard a similar statement or problem/ solution set for Graph Search. Yes, there’s a LinkedIn killer use case and a Yelp killer use case, but I’m not so sure that these will really impact these established products with loyal followers. Instead it’s in the weird connections and attributes brought together by Stupid Tricks that there’s an opportunity to create value.

It may be that Facebook will have to take the lead in surfacing interesting Graph Search data and new use cases in order to gain better adoption. Obviously there’s a great use case for advertisers, but Graph Search it comes with an Achilles heel. Advertisers are already enjoying similar benefits of Graph Search through existing Facebook advertising programs. The problem for advertisers is that unless Facebook users can find their own organic and relevant use cases for Graph Search they will likely opt out of it. And as users opt-out it will set up a feedback loop of diminishing returns for advertisers.

Facebook’s beta approach to Graph Search gives some hope that this might happen. Especially if they can be patient, build up data and let the use cases occur before setting it loose on the advertising world. There’s some other large questions, such as how relevant Graph Search will be within Facebook’s walled garden. But as an experiment that could build our understanding of how people connect, while hopefully fostering “obscurity”, I could learn to like Graph Search.