26 November 2010 at 10:32 #19361
Here’s my first blog post about TalkPhysics itself, particularly its users and usage. We have 3397 registered users and 1367 topics and comments posted as of 26/11/2010. Recently we tried a first analysis of our user data. (See bottom of this post for some caveats…)
We see that 22% of users only visited once and 50% of users log-on three or fewer times. The top 20% (315) of users log-on 10 or more times.
This is an attempt to show when users visit, which is complicated because users will have registered at different dates. The relative time is scaled according to when they registered, so 0% is the day they registered, and 100% is the day this data was downloaded.
We see that the bottom 21% of users never visited after their first day, half of users only visited during the earliest 9% of time for which they’ve been registered (e.g. earliest 16 days for someone who registered 180 days ago).
Whereas the top 19% (286 users) visited in last 20% of time for which they’ve been registered (e.g. most recent 36 days for someone who registered 180 days ago), so look like very regular visitors. This seems to match pretty well with the number of visits (above).
We see that most users log in on a Tuesday, so if you want the speediest response you could post on a Monday.
We see that most (89%) of users never post. That is what is expected from other Internet communities.
At the other extreme, we see that 80% of posts are made by only 28 users: the 1.9% of our those who posted 10 or more times. Both of these last graphs are comparable with the putative “90-9-1 rule” for Wikis, in which approximately 90% of users only view content, 9% edit content, and 1% create new content.
There are some problems with this data, so this is based on the 1506 user records we are certain are complete. The snapshot was on 12/11/2010. For the fully geeky, I took a quick look at whether the number of posts per user follows a power law:
Whilst not a rigorous analysis, we certainly seem to have a reasonable fit, as expected. We find that the probability of a user posting n times is proportional to n^-1.8. The number of visits follows a very similar pattern.
You must be logged in to reply to this topic.