Twitter apparently has a notoriously low retention rate (according to Nielsen's report from last April). At least, it has it compared to other social networks, like Facebook and MySpace (the image above).
I am suspicious.
First, there's a reasonable doubt that Twitter is a social network. It's closer to a combination of the niche broadcasting model + personal communication than to a social network. As a matter of fact, Twitter may be its own, previously unseen, thing. Why compare it to other social networks then? Maybe (very likely, in fact) Twitter is nothing like them.
It's kind of weird then to put it in the same category as Facebook and MySpace. It's even weirder to make conclusions about its potential for the long-term growth. Why? Well, mostly because Twitter did not, and probably will not, grow according to the same dynamics.
Social networks grow by adding new nodes, and by adding new connections between nodes. In short, they become bigger and more dense. This size + density combo has a specific network structure: a) random networks, b) heterogeneous random networks (small worlds), c) networks with consolidated group structure, d) networks with integrated group structure). This structure, in turn, determines who/when triggers cascades; cascades' size; dynamics of information spread; and its reach and limits. And all of this can help us assess some network's growth as the function of time.Now, the fact that Twitter does not grow at the same rate like other social networks does not mean (only) that it's retention rate is low. Its atypical growth may in fact give us a reason to think that Twitter is different.
People who signed up for Twitter back when it was founded probably found it super-boring back then. It's not only that it's no fun to follow 3 people and be followed by five, it's just that Twitter - being the open system as it is (just think all those third party-created Twitter apps) - simply took time to assume its form This is to say that Twitter, as a system, is characterized by a really high latency (defined here as "a measure of time delay experienced in a system, the precise definition of which depends on the system and the time being measured"). People who used Twitter at a beginning gave it up for a while, and then came back. It just takes time for people to get to use it.
So perhaps exclusively focusing on the low retention rate does not make much sense. Or, at least, it does not give us the full picture. But a combination of a retention rate + latency might: I would like to see data on how many people, after being inactive for a period of time, come back - not on month to month basis, as Nielsen's study did, but in a longer time frame.
Anyway, this just crossed my mind.
