Facebook's social graph

July 31, 2009

The social graph on Facebook has been the company's biggest asset, but over time it has become perhaps its biggest liability as well.

When users want to find their friends online, they think of Facebook first. For many users, "Facebook" is nearly synonymous with "social networking". They wouldn't think of using any other "social" service because, after all, their friends are all on Facebook. As far as the social networking industry is concerned, this dedication constitutes a massive customer lock-in, because no matter how much better you can make a social application, you'll start off not only without the preestablished connections enjoyed by Facebook; you'll also be fighting against the reluctance of Facebook users to try an application outside of the Facebook ecosystem in uncharted territory where most of their friends do not exist.

The Facebook developer platform (which includes the ability to write widget-like applications for placement on Facebook.com, as well as the ability to extract data about users for integration into applications on other domains) narrows this gap only slightly. For all of Facebook's talk about wanting to open up, its platforms APIs and policies empower third-party developers with only so much data and user access. Compared to the power that Facebook wields as chief overseer of its data and users, outside developers can query just a sliver of its social graph. And of that sliver, they can only store certain data in certain ways for certain periods of time. The restrictions add up so that Facebook integration delivers but minor, complementary benefits to most third-party sites.

To break things down a bit, the platform can be divided into push and pull components. Many of the APIs are designed to let you pull data about Facebook's users and leverage that data in your applications. Others are designed to let you push data from your application back to Facebook, usually for sharing user activity with friends there. These push mechanisms are the most critical for most third-party developers, because users want to retain contact with their Facebook friends and share activity with them. The data you pull from Facebook about users is generally less interesting, if only because it's pretty generic. Unfortunately, the push mechanisms are pretty weak since they don't let you reliably send data to individual friends of users, whether through Facebook's proprietary messaging system or email notifications. Your best bet is to rather bluntly dump something into the homepage stream and pray that it catches enough friends' eyes to make an impact.

All of this is to say that Facebook still has a huge competitive advantage over other social networking companies (whether on-platform or off) because it controls a valuable social graph – and particularly the email addresses that come along with it. However, the social graph is not a divinely produced thing. And it's not a permanent, exclusive good. On the contrary, I believe the social graph is deteriorating on Facebook and starting to be reproduced elsewhere in better form.

The main problem is that people's real-world social graphs change often and automatically, while their virtual representations on Facebook change mostly uni-directionally and manually. In other words, friends come and go in real life; but on Facebook, they usually just come. Friend lists tend to get bloated over time because users have a harder time defriending each other virtually than in real life. And even if they are going to defriend each other virtually, it has to be a deliberative effort, unlike in real-life when you just stop seeing certain people.

This problem is particularly acute for Facebook, because its earliest adopters were college students or high school students who have undergone significant changes in their lives over the last few years. They no longer see many of the people who they once friended in school. And they aren't inclined to remove these friendships from Facebook because they're lazy, fatigued or simply too polite.

The ill effects of this discrepancy would have been tempered had Facebook stuck to its original value proposition of static profiles. However, Facebook has undergone a major shift from a static directory to a dynamic communication channel. This shift is embodied by its decision to remake its homepage into a Twitter-like stream of directly published content. When you open up Facebook these days, you're bombarded with little bits of information about your Facebook friends' lives. It's no longer primarily a place to browse people's profiles (and associated photos) like Wikipedia pages.

Don't get me wrong, I love the "real-time web" as spawned by Twitter and advanced by FriendFeed. But Facebook has hoisted this dynamic paradigm onto a user base that didn't expect it, didn't ask for it, didn't prepare for it, and perhaps doesn't want it.

I've already discussed why this last factor is such an issue. But assuming the idea of micro-sharing does grow on Facebook users, they haven't established the right audiences for it. Friendships haven't been made on the basis of content consumption; they were made first to simply acknowledge your friends and later to gain access to their profiles (once Facebook opened up for non-students and became a less trusting environment). Sure, the news feed was introduced rather early on and aggregated information about those who users decided to friend. But the inability to post content directly and immediately to all of your friends' news feeds created an important sense of distance between you and them – and made it easier to coexist on the site with those friends who weren't really your friends anymore, or those who you didn't ever care to hear from much.

As a content producer, my predefined social graph on Facebook makes me reluctant to publish there, because I don't feel as though my friends have indicated an interest to see my constant updates. The problem I have as a content consumer is just the flip-side: when I load up Facebook, I see content produced by people who I don't particularly want to hear about or from.

Facebook has provided various ways to sort friends into lists and hide individuals from your stream, but these tools are daunting and perhaps ultimately futile. I spent 20 minutes alone last night organizing just my friends with first names that start with letters A-C. With almost 800 friends, I'm reluctant to keep going. And I imagine that most Facebook users don't even have the wherewithal to try in the first place.

Facebook may try to address this content audience problem by introducing a Twitter-like follower model. The site already asks you when friending someone new whether you want to see that person's updates in your home stream. But users won't be doing this retroactively, and it adds complexity to an already complex site. Privacy and distribution controls simply aren't going to solve the problems of an over-encompassing social graph.

What does this all mean? Well, Facebook's golden goose (the social graph) may not be so golden after all. It changes as users change. And it's not really even a singular thing. People have multiple social graphs; Facebook just tries to roughly represent them all by clumping them together. When it comes to profile access, you may want to leverage a different set of connections than when it comes to status message streams. Facebook may have to make a decision as to which particular social graph it wants to represent for its (constantly growing and diversifying) user base. It may not work for the company to be all things social for all people.

It also means that there's a massive opportunity for other social sites to give Facebook users a fresh start with fresh new social connections. I'm biased here, of course, since I'm working on social software. But this opportunity is seen in the rise of Twitter, which can attribute much of its success to the mere fact that it's not Facebook. When you sign up for Twitter, you can determine anew who you care about - whether that's your new friends or coworkers, or celebrities, businesses and media outlets. Facebook will no doubt remain a dominant social network for quite sometime, but it's dominance does not preclude the rise of other, independent social applications and services.