Facebook let strangers run its growth experiments

I argued last spring that MySpace had the registered accounts and Facebook had the people who came back, and that the published rankings would invert within about two years. Ten months on it is happening faster than I expected, and the mechanism turned out to be one I did not anticipate.

What has driven the acceleration is not retention. It is that Facebook let other people build on top of it.

Opening the platform to outside developers last May was a genuinely unusual decision. Most companies in that position guard the relationship with the user jealously, on the reasonable grounds that the relationship is the asset. Facebook instead handed strangers access to the social graph and let them build whatever they liked, and what they built was mostly rubbish. Quizzes. Virtual gifts. A thing where you bite people.

The rubbish worked, because each of those applications spreads by asking you to involve your friends, and every invitation is a reason for somebody to come back.

So the site now gets thousands of small growth experiments run on it every week, at no cost, by people motivated to make them spread. No internal team could produce that volume of attempts. It is the single cleverest thing anyone in this business has done in the last five years and I do not think it was fully intended.

It also has an obvious cost, which arrived on schedule in November. The advertising system that pushed your purchases from other websites into your friends’ feeds was switched on without asking properly, and the reaction was severe enough that it had to be walked back publicly within weeks.

That episode is worth holding onto, because the underlying tension does not resolve. The value of the graph to an advertiser comes precisely from it being detailed and personal. The value of the graph to a user depends on it feeling private, or at least bounded. Every step toward the first devalues the second, and the company has to keep taking those steps because there is no other obvious way to earn from this.

Meanwhile Twitter is being written about far out of proportion to its size. A million or so people, most of them in technology or journalism, which is exactly the population that produces coverage. I would not read much into the volume of articles.

What is worth watching there is the asymmetry. You can follow someone who does not follow you, which makes it a broadcast tool wearing the clothes of a social network, and broadcast tools have a habit of becoming infrastructure that nobody planned for.