Although there were many social networks before Facebook, the dominance of Facebook in the cultural zeitgeist cannot be denied. For many people, social network and Facebook are synonymous, especially following the release of the movie The Social Network. Of course, social networks as constructs of people go back to prehistory. However, research into the analytic side of social networks began in the 1930s, and led to the field of study called Social Network Analysis (SNA). As an aside, I’ve been working in SNA since 2000, and it is a constant source of annoyance when people assume I am just following on Facebook’s coattails. SNA has been used to study a wide variety of social concepts such as terrorist networks, improving the efficiency of communication in mergers, understanding communication blockages in large companies, dating compatibility, espionage, counter-espionage, music recommendation, gang loyalties and tracing the spread of sexually transmitted diseases.
A social network graph looks something like this:
This graph is colour-coded to indicate how well someone is connected within a social group, but there are almost unlimited ways these kinds of graphs can be analysed. At BeehiveID, we use graph complexity as one of many measures of whether a social account is backed by a real person or not. Our graphs look something like this:
In the “genuine user” graph, you can see a periphery of tiny dots that represent friends who are not connected to the core user (me, in this example). These are people who I have friended, but never interacted with. This is typical for social networks, both real and electronic. The British anthropologist Robin Dunbar posited that humans can only maintain around 150 stable relationships — this has been called Dunbar’s Number. This number has been validated in real-life social networks, online social networks, companies, military units, etc. So, Facebook “friends” are not really friends in the Dunbar sense — they are more accurately “connections,” a loose conglomeration of real friends, co-workers, past teachers, family, extended family, old schoolmates you don’t remember, etc.
Facebook is one of many social networks, some of which are more social than others. Facebook is the behemoth, with a staggering 1.1bn monthly users, as of mid-2013:
The average Facebook user has around 338 friends, with a median of 200. At BeehiveID, we have seen a number of users with more than 5,000. So clearly people are connecting with more people than they can actually maintain relationships with. Other networks include:
This is just a sampling of the large US networks. The Chinese social network RenRen has about 220m monthly users. These are massive networks of people, and the data associated with them is tantalising to SNA researchers like myself. Some of my colleagues complain that it isn’t “fair” that they should have so much valuable data that is unavailable to researchers. I can only imagine what cool views the data scientists have at these large companies. However, very little other than the most high-level statistics ever escape these walled gardens.
When these networks were created, most had application programmer interfaces (APIs) to allow outsiders to do some limited analysis on the data behind the walls. Usually, this was in the form of the OAUTH protocol, which allows a person to authorise an app to look at their individual social network. This was limited to 1st level connections only, though. In other words, you can only look at yourself and your friends. But it was still valuable data to examine; perhaps too valuable. In April of 2013, Facebook severely restricted their API, eliminating the access to 1st level connections, essentially killing the graph part of the social graph. I’ve written about this, and it caused us quite a few problems at BeehiveID. In February 2014, LinkedIn followed suit, eliminating access to 1st level connections to a very small number of trusted partners. This restriction was inevitable. Many so-called “data harvesters” were abusing these APIs to gather as much data as possible. But many legitimate apps have been caught in the crossfire.
The impact to dating sites and other internet properties has been dramatic. Many dating sites relied upon friend (or friend of a friend) recommendations for dates, and that has been eliminated totally. One notable exception is Tinder, which doesn’t rely upon the social graph — it uses individual Facebook login and matches you to people in your geographical area. Sites which were dependent upon the social graph of these major properties have had to adapt other methods or go out of business.
Of course, this raises the question of why these sites ever granted access to their data in the first place. It always seemed odd that a site whose data was so valuable would open up an API to let anyone access it. I don’t know the actual reason, but I can speculate it was mostly tied to when these sites were created. At that time, APIs were all the rage, and fostering a welcoming developer community was seen as the path to success, or at least free buzz. However, I think developers are extremely skeptical of free APIs now.
Case in point: in an article in TechCrunch about F8, the Facebook developer conference, Josh Constine writes the following:
“Keeping partners loyal and feeling valued has been sore spot for Facebook in the past. It cratered game companies like Zynga when it reduced virality on its web platform. News reader apps from The Guardian and Wall Street Journal got burned when the News Feed stopped promoting them. Facebook has copied features of apps built atop it. Platform whiplash rocked developers as Facebook rapidly changed its APIs around 2009. Facebook’s little-known HTML5 app platform was abandoned soon after birth. Now, Page admins are up in arms as their News Feed reach decreases.”
What is the answer? Basically, you cannot depend upon free, unfettered access to someone else’s data. If you get your users through Facebook, they aren’t your users. You will have to build a community or your social graph the hard way – by advertising, promoting viral growth, incentives, referrals, etc. – the traditional methods used before these huge social networks came around. Online social networks were a convenient jump start, but in the long run, owning your own data will be much more valuable to your company and your users.