What is meta? ×
Meta Stack Exchange is where users like you discuss bugs, features, and support issues that affect the software powering all 127 Stack Exchange communities.

I was playing around with data.stackexchange, and I decided to see how many users were actually active on Stack Exchange. I made this data query, and these are the results for a selected number of sites.

+--------+------------------------------+---------------------------+
|  Site  | % who posted in last 30 days | % visited in last 30 days |
+--------+------------------------------+---------------------------+
| Math   | 4.6%                         | 26%                       |
| SO     | 4.1%                         | 32%                       |
| TEX    | 3.0%                         | 30%                       |
| WP     | 3.0%                         | 24%                       |
| Gaming | 2.2%                         | 24%                       |
| Ubuntu | 2.0%                         | 26%                       |
| SF     | 1.4%                         | 22%                       |
| UX     | 1.3%                         | 31%                       |
| SU     | 1.2%                         | 22%                       |
+--------+------------------------------+---------------------------+

Here is the actual Google Docs Spreadsheet. Feel free to edit and improve it.

Looking at that data, it seems that a lot of people are creating accounts on Stack Exchange, and then leaving.

I think that we should try to get some of these people to return: it would be a more efficient way of promoting the site than trying to get new people to join. Or, at the very least, we should find out why they left, and use that information to improve Stack Exchange.

share|improve this question
24  
I think at 20-30% return rate is pretty good actually, especially as you've counted users over all time. –  ben is uǝq backwards Mar 18 '12 at 20:45
2  
Expanding on Ben's comment it would be interesting to see how these figure depending on time-since-account-creation and how the cumulative figures were trending. –  dmckee Mar 18 '12 at 20:54
    
@Ben yes, but that doesn't mean that the 70% of users couldn't be useful. –  Christofian Mar 18 '12 at 20:56
    
+1 @dmckee, yes it would. My account says I've been a member for 1.5 yrs but I've been "active" for far less time than that. Christofian, I agree, but I think the bigger gap is the posted vs visited. You won't rescue everyone but the people who use SE are more likely to be persuaded. –  ben is uǝq backwards Mar 18 '12 at 21:09
    
@Ben check the spreadsheet: I added a column that shows the users who posted in the last 30 days / the users who visited in the last 30 days. –  Christofian Mar 18 '12 at 21:45
10  
Yeah, those are pretty good retention rates. Most people excrete bricks when they first see the bounce rates/retention rates for any given site. It's a big, big world and a great deal of people try something once and stop using it. –  Ben Brocka Mar 19 '12 at 0:51
1  
Too bad we have no idea what those numbers are for Beta sites... –  Ivo Flipse Mar 19 '12 at 11:20
add comment

1 Answer

Quick aside, don't use GET[UTC]DATE() like that. Data dumps are snapshots in time, so using today's date isn't really meaningful; hard code the restore date or select a max date out from a relevant table.


You probably want to be more discriminating when looking at returning user statistics, not all users are equal contributors to the health of a site.

For example, users who posts answers would be missed more than those asking them. While both tasks are necessary, answerers are the ones actually producing the re-usable nuggets of knowledge that Stack Exchanges are striving to produce.

You probably also want to exclude low quality content, we care far more about finding and retaining good contributors than a raw head count. A post's score being positive (> 0, that is) is a good proxy for "quality content", though not perfect obviously.

Some rate of attrition is expected as well, people get jobs (if students), change to more demanding jobs, get married, have kids, and what not. All these things will naturally cause old members to fade away, independent of whether they're bored, angry, tired, etc. of Stack Exchange.

All that being said, I threw together a query that captures some of what I've said. Mostly in the "quality contributors" and "Q/A divide" sense.

Results on Ask Ubuntu for comparison*

PercentAllActive PercentAskersActive PercentRecentAskersActive PercentAnswerersActive PercentRecentAnswerersActive 

---------------- ------------------- ------------------------- ---------------------- ---------------------------- 

27.7%            32.0%               55.4%                     38.4%                  72.0%

*Active here is "visited the site based on LastAccessDate".

share|improve this answer
    
I'm new to SQL queries, so mine wasn't that accurate. Thanks for improving it. That said, yes, some people move on, and some people post bad content, but I think it still would be useful to see if we can see if we can get some of the people who have made contributions but who are now inactive to join the site again. It least, it would be easier than getting completely new users to join. –  Christofian Mar 19 '12 at 1:00
    
Interesting assessment. I've always been of the mindset that the question askers are the more valuable. Each great question could give birth to 3 or 4 really great answers. Without the questions, the answers just won't exist. We see this manifest occasionally when a helpful edit to a NC/NARQ question turns it into something great that gets amazing answers. Without that edit, those answers wouldn't be, and the question would just be a closed, empty shell with no answers. –  jmort253 Sep 29 '12 at 20:06
    
Kevin, will the dumps be updated soon? It's been nearly 3 mo since the last one... –  Lorem Ipsum Sep 29 '12 at 20:50
add comment

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .