15

The Stack Exchange Sites list provides a list of all of the sites and some statistics about them: total questions, total answers, percent answered, users, visits/day, questions/day, and age.

These probably aren't the best metrics. For example, total questions and answers don't necessarily mean a whole lot to me. In most cases, I'm searching the site for the information I need. The total number of users is also relatively uninteresting, although a total number of active users (probably timeboxed and cached) would be more interesting.

Jon Ericson wrote a Stack Exchange Data Explorer query to report on the number of questions, median views, average score, closed rate, average answers, acceptance rate, answer rate, and median time-to-answer for a given site. Not all of these are appropriate for the Stack Exchange /sites list, either.

There are two aspects to this:

First, the /sites list should be updated. I'm not entirely sure what metrics I would display. I think that answered percentage is useful, as would be visits/day, questions/day, and site age from the current lists. Median time-to-answer from Jon's query would also be interesting.

Second, every site should have additional stats available. These should be cached, but storing things like total questions asked, total answers, percent answered, total accounts, active accounts, visits, questions, age, average answers, median views, median time-to-answer would be nice. These would be a whole lot more detailed and specific to one site.

2

I'm not at all sure what other people use this list for. I mostly look at questions per day, which turns out to be a useful measure of graduation-readiness. Before a new site enters beta, I sort by name to make sure we aren't already using the initials for the favicon. (That hasn't prevented History and Health from shating an H.) Occationally, it's helpful for me to compare visits per day. But these seem rather specific to my role as a CM.

I suspect the main purpose of the site list is to help people find interesting sites they might like to participate in. But I bet the Hot Network Questions sidebar does a much better job of attracting cross-site participation. Anyway, all of this is to say we should probably figure out what the page is for before redesigning it.


I originally wrote the query to self-answer Does it pay to spin off sites? It used a group of metrics suggested by this paper. For the most part, the statistics don't really compare well across sites. Median time-to-answer, for instance, ranges from 22 minutes (Math) to nearly a day (Health). All things being equal, faster is better. However, all things are not equal. Smaller sites tend to have longer wait times and it just takes longer to write a useful Philosophy answer than it does a Unix answer. That's not an argument for including it or not, however. Just something I've noticed.

To get an idea of how that stats look for your favorite site, I've put the question stats so far in 2016 in this spreadsheet. For some reason, I can't get Stack Overflow's stats as the query now times out however short a timeframe I pick. (I probably should start using R to do medians and other statistical measurements.) In any case, I find the per tag statistics more meaningful.

For the more detailed site stats you suggested, I wonder if some of the statistics would be better presented in a timeline form? We could reuse the framework of the site analytics page, perhaps. While I'm thinking about it, I've been meaning to check back on that privilege to see if:

  1. The page is ever used, and
  2. The privilege is something people actually pursue.

If the page isn't working well as an incentive, perhaps we should repurpose it for public interest.

  • 1
    It's not just CMs. When I was active on beta sites, we'd frequently use that list to try and see how well we were doing. Since the graduation guidance was vague and subjective, it was somewhat useful for e.g. the Japanese site to monitor and compare stats with the German and French sites that entered beta around the same time to try and figure out if we were more on the "problem" end or "doing okay" end of the scale. Specific values I used were total questions, questions/day, visits/day. – Troyen Jun 3 '16 at 20:39
  • The site analytics page might've been useful, but the rep threshold is too high for 95% of the people I remember interacting with on the beta meta site (largely in the 1k-3k range), so we resorted to public trackers of Area 51 information. – Troyen Jun 3 '16 at 20:43

You must log in to answer this question.

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