If you want to have staying power on the Internet, you need to have turnover, says a new analysis published in the Proceedings of the 2010 ACM conference on Computer supported cooperative work.

Not only do you need to be 'heterogeneous', you need to be diverse.   

Folks, we may be in trouble.   We only write about science and no one ever leaves which means we are doomed to fail.   Obviously you are wondering how it is possible to even study such a thing as an online community.   Something like 4Chan should not work at all, because it is complete chaos and a leftist Internet hate machine and everyone knows Republicans control all the media.  No study could have predicte 4Chan or there would be a lot of rich professors out there.

But sometimes you have to pick a methodology and take your shot.   Daphne Raban of the University of Haifa, along with  Mihai Moldovan and Quentin Jones of the New Jersey Institute of Technology, did just that and set out to dispel belief that online communities could not be effectively studied so instead of focusing just on group size and activity, they established a metric with social characteristics, such as the group's homogeneity and heterogeneity - homogeneous when member turnover is small and the members who established the group are still the main members, while if the group has turnover and new members are continuously joining it, it is heterogeneous.   Whew.  New people join every day so we may be out of the woods.

Here is the tricky part - the method.   They sampled 282 chat channels "born" in the same month - born being when at least three members had exchanged at least four messages in 20 minutes - and then followed them for 6 months and did a survival analysis.   A channel was considered "dead" when it had no activity for four weeks.

They examined activity two hours after 'birth', the first day of activity, the first week of activity and the first two weeks of activity.  What they gathered was that member turnover was the key thing, not how many people posted comments or how many actual members it might have.  But turnover must mean different things to sociologists than it does to you and me.  I think of turnover as a corporate suit does - people lost and people replacing them.   They only mean people gained.

So here is our 'turnover for the last three weeks:



So new people are 'in' every day but were there any people 'out'?   I have no idea.  Because people are not contracted writers they come and go and never actually 'leave'  - someone could be gone for a year and not write and that is no issue and when they come back they write just like they were here yesterday.

It can't tell an outsider anything about our ability to survive, since most new members do not contribute content or even comment, membership just makes the articles and blogs easier to track and allows chat and other social things anonymous people cannot access.  Chat rooms may be more valid in that kind of study since the barrier to entry is much lower, though 4 weeks to be declared 'dead' seems awfully long.

What might explain our success is the number of messages that are posted between members, which the researchers also regard as a good barometer - we are not talking at the audience, we are also part of the audience, and their analysis stated that the number of messages between members correlated to higher chances or survival.

But if the ratio between the number of messages and the number of members remains the same after two weeks of the community's activity, the chances of "death" become higher.   Irregularity helps, it seems, and neither increasing ratio of messages between members nor decreasing made much difference. 

"Prediction of an online community's survival chances cannot be based on quantitative data relating to the size of the group or even to its growth rate alone. A social predictor, on the other hand, can much better predict its chances," said Raban.

So we may never be Facebook size but we aren't doing too badly either.