Both, Area 51 and stackexchange.com display the total number of users as one of the statistics available for each site. Area 51 additionally displays the number of avid users defined by meeting a certain threshold of reputation on the site. There is also a graph of new users in the moderator tools.

I consider the total number of users, as it is displayed currently, a completely useless statistic. I have no way to determine how many of those users are actually active on the site, I can only see how many people have registered. Many of those might never have come back to the site. The number of new users in the moderator tools is also not very useful, as I have no idea if those new users actually come back to the site and contribute to it.

The number of avid users on Area 51 is also not a sufficient statistic, as it doesn't tell us if those high reputation users are still contributing to the site. Most users don't rage-quit and have their account deleted, they just fade away silently. Most sites likely have some users with a lot of reputation that are not contributing anymore.

The most important thing I want to see from the user statistics is whether the site is gaining or losing active users. A site that is steadily gaining users that actively participate in the site, either by contributing posts or maybe just by reading and voting on the content, is a site I would consider healthy (barring any other severe problems). If a site that is hemorrhaging users, if the number of active users is shrinking over time that would be a very clear warning sign that there are severe problems with the site.

If the user base of a site is shrinking, it would be useful to see this problem as soon as possible to be able to take countermeasures. This is where a better statistic for the number of active users would be very handy. This would be especially important on beta sites, to better gauge the health of the site.

For this statistic we need to define what constitutes an "active user", some ideas would be

  • Has visited the site in the last 30 days
  • Has performed at least x actions in the last 30 days (actions would be new posts or edits, maybe also voting and commenting)
  • Was active at least once before (to filter out new users, this statistic should show the people who stick around)

I've create a very basic SEDE query that shows the number of active users (posted at least once in the last 30 days) and here is another query from Sklivvz. They give very different results, probably some error in my query. Though more interesting would be the time-dependence of that number, but my SQL knowledge is not sufficient for that query.

One might also split the statistic into active users and passive users, who only read the site, but do so regularly.

This statistic would be most interesting for beta sites, so a SEDE query would be not possible. Integration with the moderator tools, so that the mods could see the number of active users over time would be most useful.

share|improve this question
    
The criteria for an "active user" are just some very rough ideas, if someone has a better idea please post it here. –  Mad Scientist Aug 20 '11 at 17:39
7  
we are aware of this, and it ties into a "citizenship rating" we are considering to replace "accept rate", etc. But no ETA at this time. –  Jeff Atwood Sep 12 '11 at 6:55
    
Possibly better query: data.stackexchange.com/stackoverflow/q/112331/… –  Sklivvz Sep 12 '11 at 7:55
add comment

4 Answers

Here's an idea for a metric:

  • Visited the site within 14 days
  • Earned reputation within that period (upvotes, accepts, approved edits)
  • Performed any "new" actions (edits, new answers/questions) within that time

My reasons are:

  • Active within a month is a pretty lenient metric. Two weeks shows, more or less, users you could reasonably expect to be around within the next few days.
  • Requiring earned rep shows constructive participants, and only counting recent rep means they have recently done something constructive approved of by other users. Association bonus should not count. All other positive rep gains are caused by community members manually approving of something you have done.
  • Checking the site periodically and gaining rep slowly on old posts doesn't make you active, so the third requirement balances out the earned-rep one

If the site's users are leaving or not participating as much, this metric really should show it.

share|improve this answer
    
I'm not sure why 14 days is better than 30 (except that it's a shorter period). I just came back from a 2 1/2 week vacation and now I'm catching up on sites that I consider myself active in. Also, I think there's a real value in having people who read a site, vote on posts, and nothing else. I wonder if rather than one, be-all-and-end-all active user stat, we break it down into active readers, voters, askers, answers, and so on. –  Jon Ericson Jun 28 '12 at 4:29
    
@JonEricson but you weren't there for two weeks. You weren't active at the time. Now you're doing stuff again, so you are That's the point. In addition, a site full of people who post once a month is hardly active. Voting is important, but it's not a very useful metric on it's own; visitors don't come back to the site to see new votes (not non-registered ones anyway), they come back to see new content. –  Ben Brocka Jun 28 '12 at 13:05
    
Hmmm... I think we have different uses in mind for the statistic. What I'm looking for is an measurement of a site's user base. I'm involved in a lot of sites where I mostly read the questions and answers, but not contribute unless there's something I have real knowledge to add to what's already there. True, I'm not active in creating content, but I could be if given the right stimulus. The reputation league page gives me, more or less, the stats you suggest. –  Jon Ericson Jun 28 '12 at 18:56
add comment

Strong agreement.

At Writers.SE, even though our "Avid Users" score is "Excellent" at 206, I'm well aware that our actual number of active participants is far, far lower than that. As a mod, I know how many notable regulars we have, but I have no reasonable way to estimate whether we've got lurking readers, consistent infrequent contributors, etc.

User activity is a crucial stat for a beta site. Even a simplistic query, such as "number of distinct users who posted and/or voted in the past 30 days," would be far clearer and more indicative than the monotonically non-decreasing "number of users whose rep is above 200." The current statistic is certainly significant in its own right, but some sort of metric describing the scope of the userbase is definitely worth publicizing as well.

share|improve this answer
add comment

There is a "traffic" statistic that tracks the number of visitor sessions per day, but from a description of how it is calculated it looks like users can be counted multiple times if they visit several times in a day. While better than simply counting the number of people that have registered, the traffic stat isn't good enough because a small group of super-active people can make the site's numbers look higher than they really are.

As an example of why the current metrics are insufficient, someone on the Japanese Language and Usage beta site started tracking the public Area 51 metrics and plotting them on a graph: Graph of JLU users over time

The number of users (yellow) has been increasing constantly, but the number of "avid users" (blue) has remained almost flat. This suggests that the new users aren't really sticking around. Most of them probably register an account, maybe ask a question or two, and then vanish. My anecdotal memory of question askers backs this up - I tend to see the same set of names pretty frequently, and only rarely notice a new name or two.

The traffic (red) on the JLU site has been slowly trending upward with several spikes that correspond to a flurry of activity. Here's the questions per day graph for comparison: Graph of JLU questions per day

Despite occasional spikes in activity and questions, the active user line receives no corresponding bump. Combined with my own anecdotal evidence again, this suggests that our traffic spikes are from the existing avid users.

So, none of the public Area 51 statistics really measure how well a community is doing, except in the most general terms. You can try to infer it by checking the reputation leagues and voter lists at the end of each month, but even that method is imperfect.

An actual "active user" statistic publicly provided by the site would make it easier to accurately measure the health of a beta community over time. The currently available metrics can be misleading and overestimate the activity of the userbase.

share|improve this answer
add comment

I fully support this request. Since the "avid user" stats on Area51 only look at rep threshold and you get 100/200 points for free for associating an account, sites usually end up with several users with 200+ rep (which you can easily get from two questions/answers in a private beta), who have all but stopped participating on the site.

I noticed this trend on gardening.SE, where Area51 lists 89 users as "avid", but only a handful actively contributed to the site. So I dug up some info using the API (that's the only way to access info for beta sites) and compiled some stats. Only about 2.4% (13 users) of the user base (533 users) accounted for more than 1/2, 2/3 and 1/4 of upvotes, answers and questions respectively. Looking at the "last login" date on the "avid users", I could see that a good number of them hadn't logged in for a while (this info wasn't shared in the meta post).

While I was not completely shocked by the results, it certainly was very illuminating and definitely something that would be nice to have in the mod analytics page. It helps to visualize quickly if the active user base is growing or shrinking or staying flat and how much of activity is generated by them.

share|improve this answer
3  
Don't you need it to be Seasonally adjusted, for gardening.se... –  Ian Ringrose Sep 5 '11 at 14:29
    
@IanRingrose We've only been up for 3 months now, and we'd need more data to look for seasonal trends. Right now, I can only say that people were initially active and buzzing during the first 3 weeks of beta and then stopped. However, I'm certain that seasons will play a part and I'm seeing a bit of that right now, as traffic has been steadily increasing for the last 10 days, because it is lawn season in most places. –  Lorem Ipsum Sep 5 '11 at 14:49
add comment

You must log in to answer this question.

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