Some challenges for Facebook’s Graph Search


Facebook’s upcoming Graph Search feature has already been plenty hyped, even though its weeks away from a significant portion of the public being able to try it out. But in order for Open Graph to be meaningful, it’s going to have to tackle some significant barriers.

Let’s take one of it’s core promises – that users will have powerful search tools for local businesses available to them, by tapping into data about what one’s friends have Liked. Ostensibly, a user could utilize the recommendations of his friends to, say, find a good restaurant in a town he’s visiting. “Social curation” is the idea. But in order for this to work, let’s break down what has to happen first.

Let’s say I plan to visit Atlanta and want to find a suitable place for dinner. I have a handful of friends there (I live in KY), so this is a viable scenario. I search for “Restaurants my friends in Atlanta have Liked”. In an idealized world, I’d have a nice little list of places my friends have visited, and a few moments later, I’d have made a reservation and moved on.

OK, but not so fast. How many decent recommendations am I likely to get? I’m not sure it will be many, if I look at the filtering that has to happen first:


I’m going to do some back of the envelope math here. I don’t have any data on this, so this is just pure guesswork. Let’s say I have 30 friends who live in Atlanta. 5 are ultra concerned about privacy, so they’ve kept most of their Like data private. Another 12 use Facebook regularly, but don’t care about Liking or Checking In every place they visit. I can’t expect decent data from them. Down to 13. Let’s say that 7 of the remainder don’t really share my tastes – they’re acquaintances, in-laws, or people who simply enjoy different things that I do. OK, so I’m now left with all of 6 people who have Liked restaurants on Facebook in Atlanta. But I’m not done yet – Atlanta is a big place, if a recommendation is across town, then forget it. And what if I have a recommendation for a fast-food cajun place nearby, but I’m actually looking for something upscale?

You get the idea. Even though the amount of information Facebook has in aggregate is mind-bogglingly massive, the amount of useful data that an individual user can obtain through her friends is quite different. Presented with the scenario above in real live, I’m still going to turn to something like Yelp, Urban Spoon, or TripAdvisor.

All of this isn’t to say that Graph Search won’t work or be useful. It could be tremendously useful. But in order for that to happen, Facebook is going to have to provide users with more of a reason to share data about where they visit and what they do. Otherwise, I’ll take the recommendation of 500 strangers over 5 friends any day.