The Problem With Facebook Data
The more I use Facebook Graph Search, the more evident it becomes that Facebook made a major mistake with their most ubiquitous feature: the “Like.”
Follow this train: Facebook’s value, their kryptonite, is their data. They have a billion users, and they know the habits and preferences of all of billion of them because they can easily track that information by examining what they’ve Liked.
Or can they?
On the pre-Like Facebook, users were Fans of pages. That information – which brands, bands, books, movies, sports teams, etc. a user was a Fan of was prominently displayed on the user’s profile page. Which meant users spent a lot of time curating those selections, pruning and adding so that the list was an accurate reflection of who they were. Or at least who they wanted people to think they were.
As a result, it was tough (or tough-ish) to get users to become fans of pages they didn’t think would give them social currency or look good on their wall. Hence, the Like, an easy way to give a brand a thumbs-up (and permission to coat your wall with brand messages.)
But while the Like button has become ubiquitous and a seeming smash hit for Facebook, it does not appear to be used in any consistent manner. That was its selling point: a lower key way for users to indicate approval for a brand, but it’s also it’s Achille’s heel: if users aren’t displaying any sort of consistency in the way they use the like button, then the resulting data is fairly inaccurate and not all that useful. (Bye-bye monetization.)
This is evident in the spate of Tumblr blogs flagging the random overlaps Facebook Graph Search pulls up (married men who like prostitutes, Christians who like porn) and in less quirky uses, like the study BTIG did on the accuracy of using Graph Search as a movie recommendation engine. What they found should be somewhat troubling for Facebook: the Likes users entered dated back to when they first signed up for the service or last used the Fan format.
There are multiple reasons users are either promiscuous or inconsistent with their Likes, but they all circle around two competing forces: (A) If The Brands We Like are how we present ourselves to the world, it takes a lot for a new brand to crack that list and (B) if Likes are easy to give and easily buried in the News Feed, there’s no reason not to give them out at random.
Now by brands, I mean any sort of product: movies, books, songs, actors, vacation spots, along with the more typical products and services we call brands. To the consumer, they’re all a part of who they are, and attaching their name to anything other than the tried and true, when that preference is in a prominent location, is a leap.
The constant rejiggering of the Facebook interface – particularly the introduction of the Timeline, where the “About” section, with the brands and media the user likes, is now one level down– has lessened the amount of attention people give to their Likes. That in turn, works on the validity of Likes from another direction: those who don’t see their Likes as a reflection of who they are are likely to become more promiscuous with them, and at the same time more random, on the assumption that the action has both an immediate value (unlocking a coupon offer) and limited aftereffects.
The other culprit here, perhaps even more to blame than Timeline is Frictionless Sharing: by posting every article the user half-glanced at and every song they or their offspring started to listen to, Frictionless Sharing greatly reduced the impetus the user had to be sparing with their Likes. While “Mary is listening to Ruby Tuesday by the Rolling Stones on Spotify” was supposed to be of lesser significance than “Mary Liked Ruby Tuesday by the Rolling Stones,” users did not really grasp that subtlety and many likely figured that if all bets were off, why not just start Liking just about everything?
While it’s possible to imagine a scenario where Facebook encourages users to carefully parse out Likes to new products so that their opinions can help guide their friends, it seems unlikely: first and foremost there is the black and white nature of the Like: it’s an endorsement, pure and simple, in the way that 3 stars out of 5 is not.
Can Facebook fix this? Probably. They’ll need to rethink both the Like and frictionless sharing and the value users place on having their brand and media preferences prominently featured on their profiles. If they can solve for that in a way that encourages more, rather than less interaction, and more curation of the brands (media and otherwise) a user Likes, they have a good chance to make their data more accurate and thus, more valuable.
The stakes are high, and it’s a big “if.”