Tim O'Reilly has posted a very interesting analysis on the "Long Tail of Facebook applications" (chart from the report shown). He notes that it's very top-heavy, with the top 84 apps of the 5,000 analyzed having 87% of the traffic. People have rightly pointed out that this is just the opposite of Long Tail theory, which would argue for a "fat tail", with the top 100 having, say, only a bit more than half the market.
(As an aside, these market share calculations depend entirely on what you call "head" and what "tail", and in my analysis of traditional vs. online markets, I always use the inventory of the largest bricks-and-mortar retailer to define "head". That, however, isn't meaningful for something like Facebook apps, so some arbitrary division is inevitable.)
So does Tim's data challenge Long Tail theory? Well, let's review the theory. It says that:
A) The natural shape of most markets is a powerlaw (Pareto) distribution. Since powerlaws are "fat-tailed", that would tend to put a significant part of the market (usually between 25% and 50%) in the tail, although see above about definitions.
B) If your market doesn't exhibit a powerlaw shape, you're probably doing something wrong in terms of driving demand down the tail. Those problems usually fall under the category of "poor findability" (from search to recommendations), although some markets are just inherently limited in terms of choice and thus don't really fit into the theory.
The best way to test whether a market is indeed a powerlaw is to rank the items in order of popularity and plot them on a log-log scale, where it should form a straight line (the slope of the line will vary from market to market and corresponds to the exponent of the powerlaw). You can't just eyeball a chart like the above and tell if it's a powerlaw, since similar curves, such as exponentials and lognormals, look similar on a linear scale. I've asked Tim for the underlying data so I can plot it log-log and see if it is indeed a powerlaw [UPDATE: his team has now done that; it's not a powerlaw. Details here], but in the meantime I did a little calculation for fun:
The simplest powerlaw is simply 1/x, and applying that to a dataset of 5,000 apps, you would expect the top 84 to represent 55% of the total usage, with the bottom 4,916 accounting for 45%. So by that standard, the Facebook data looks very head-heavy. Which raises the possibility that it's not a powerlaw at all (and thus doesn't fit into Long Tail Theory).
Which, in turn, raises the question as to why not. Facebook does all the usual things to help people discover new apps, including search, categories, rankings by users, activity and even newest. And one presumes that the the missing tool of finability--recommendations--are addressed by the built-in social network effect of seeing what your friends are using, which amounts to a sort of playlist sharing.
So it's not obvious why the bottom 4,916 apps are faring so poorly. I can only think of two possibilities:
- The social networking on Facebook is too powerful. This is the tyranny of network effects, where viral success is the only kind and popularity snowballs into an avalanche or goes nowhere at all. That sort of herd behavior is usually a sign of an immature market.
- Most apps are total crap. That, in turn, may say something about the whole idea of Facebook as a platform. But I'll leave that discussion for another day.
- [UPDATE]: As Neil points out in the comments, there is limited screen real estate on your page for apps, so you really can't take advantage of unlimited choice. In a sense, this is a case of limited "shelf space".