What is meta? ×
Meta Stack Exchange is where users like you discuss bugs, features, and support issues that affect the software powering all 128 Stack Exchange communities.

I realize that summary statistics, such as the # of questions, # of users, % answered, # visits / unit of time, etc., are some of the metrics tracked.

After thinking about the time to answer (see this post) as one metric that may be improved, I wondered about others. I found this post that mentions some other simple statistics for tags.

Are there other metrics that are commonly discussed, either on Meta or in other places, such as J&J's blogs, the podcast, etc.? Some that come to mind:

  • Time to {answer, accept}
  • Proportion of questions labeled as duplicates -- measures negative performance of search tools, i.e. either not used or not finding related questions, and could be broken down by those cases where an asker queried before posting a question.
  • Time from query to click-through (or even to upvote) -- measuring performance of SO's search tools.
  • Proportion of questions killed by OP when clicking a link in the "Similar Questions" list (or in the Q. with similar titles) -- measures performance of similarity matching (Update: This has been proposed before. How did I discover that question? Guess. ;-))
  • Ratio of Google referrals to "native" referrals -- this is pretty open-ended.
  • Ratio of views to upvotes
  • Time to return / revisit (i.e. coming back to SO) -- Could measure engagement. With a quantified metric for a regular visitor, this could measure a conversion from one-visit newbie to regular visitor. (To keep this clean, let's not discuss addicts.) This could also be conditioned on prior actions, such as posting a question.
  • Proportion of interlinked questions (and the rate of interlinking by users) -- this could address site exposure and familiarity. When considering changes to the site, being able to enlighten users about the Q&A that already exists could be quite important, and measures of interlinked questions could reflect that.
  • Ratio of responses to views -- responses could be votes, comments, flags, etc.; this is a bit tricky because the early viewers and the later viewers are likely to be different populations, and the Q&A content itself may have changed. Still, in aggregate, the responses for the first 20 viewers could be informative of engagement trends, and viewers 100-200, 1000=5000, etc., could be useful for other engagement trends. Questions that get a lot of views but little or no response implies there's some kind of issue. It could be that the question is interesting, but not interesting enough to up/down vote, or that viewers are not engaged (e.g. they come from Google and do not convert to SO members).

Naturally, many or all of these can be tracked over time to describe the health or evolution of an entire site, of different tags, and more.

Are there other metrics that are frequently mentioned? I'm sure there are some that are much better than my stream of consciousness suggestions. :)


Note 1: For those inclined to study such metrics, some sub-communities, such as those around the [r] tag, involve users who are promoting SO as a great Q&A option. When studying and comparing tags, such communities may be interesting to examine. Needless to say, the R user base might include many who are quite interested in studying such data.

share|improve this question
add comment

1 Answer 1

There's a SEDE query (provided by Jon Seigel and Greg Hewgill as a part of the Stack Exchange Community Statistics effort) that uses the ratio of Unanswered Questions to Unanswered % Rank (UnansweredCount / UnansweredPctRank) as a rough measure of tag quality. You can compare today's results to the rankings from August to see that the tag is shooting up the ranks to compete with as the "worst tag on Stack Overflow." The query can be run for any SE site that's out of beta to see which tags could use the most attention.

share|improve this answer
    
Thanks! Those are great links. Insights within and across tag communities are definitely interesting. –  Iterator Nov 19 '11 at 13:32
add comment

You must log in to answer this question.

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