21

The SO company page got an update, including 2015 statistics. I decided to compare them against the previous version, with 2014 stats:

+------------------+-------+-------+--------+
|                  | 2014  | 2015  | change |
+------------------+-------+-------+--------+
| Visits           | 900M+ | 3.9B  | +333%  |
| Pageviews        | 6.4B  | 7.9B  |  +23%  |
| Questions        | 3.1M  | 3.7M  |  +19%  |
| Answers          | 4.5M  | 4.6M  |   +2%  |
| Edits            | 2.7M  | 5.0M  |  +85%  |
| Comments         | 17M   | 17.9M |   +5%  |
| Upvotes          | 21M   | 24.4M |  +16%  |
| Downvotes        | 3.2M  | 3.8M  |  +19%  |
| Accepted Answers | 1.8M  |       | ?      |
| SO "users"       | 3.8M  | 5.0M  |  +32%  |
+------------------+-------+-------+--------+

Which raised a few questions...

  • What happened to the visits statistics, a different way of counting?
  • Were there really about twice as many edits in 2015 as in 2014? I find this hard to believe, given that the other stats like Q/A/comments/votes have not changed much.
  • Why is the number of accepted answers omitted in 2015? Is it because it's lower than in 2014? (I wouldn't be surprised.)
  • In 2014 visits already said 900M+, it might have been close to that 3.9B.... – rene Feb 25 '16 at 9:05
  • I'm not sure if 2015 was lower than 2014 for accepted, but you're right that one issue is that the stat generally feels misleadingly low. It implies a lot of questions don't get good answers where the issue is more often that the asker didn't know enough to accept. But a better reason is that it's confusing - in some places that it's used, it means, "asker accepted" (which seems most intuitive), but in others, it means "asker accepted or an answer got enough upvotes," I think? – Jaydles Feb 25 '16 at 13:43
  • Sam can confirm here, but I believe the big problem with Accepted Answers is that we weren't really counting accepted answers - we were counting questions with upvoted answers but NOT accepted answers or something like that. IOW we did it wrong last year and probably can't easily know for sure how wrong it even was (without recounting all votes according to whatever date we happened to run it last time). – Shog9 Feb 25 '16 at 16:17
13

Yup, we updated the stats on the About page. And in the process we realized we previously misrepresented a few of the network-wide metrics.

Visits - I have no clue how we ever came up with the 900 million figure for 2014. Anyone can see on Quantcast that the network gets about 900 million visits... every 2.5 months.

Edits - It wasn't clear to me how we calculated edits for 2014. So I went simple, counting all post body edits, and used that number for 2015. You can see the query here. According to the new method, Stack Exchange had 4.3 million edits in 2014 and 5.0 million in 2015.

Accepted Answers - Indeed, we removed this metric mostly because we couldn't decide how to define it. Should we count accepted answers on questions that are deleted? What about great answers that are never accepted? What about accepted answers that are downvoted? How should we window the count (within x days after the question is asked) so we can compare one year vs. another at any time as apples to apples? Etc. I'd be curious to see how you might define it. Feel free to drop a query as a comment to this answer.

You must log in to answer this question.