15

I was writing an answer on this now-deleted question which basically asked 'Why do I get an answer quicker when I post on Stack Overflow than on Server Fault?'. This triggered me; was this just an N=1 case or is there really a pattern? Stack Overflow gets a lot more visitors, sure, but it also gets a lot more questions. That can't be the reason; the same applies to weekend vs. working day questions.

Quoting part of a comment by n8te:

I believe another factor is the nature of the questions. On Stack Overflow, if you post a quality question with the necessary MCVE, error messages you're receiving, expected output, etc, people who are experts at the language can often identify where you're going wrong almost immediately. With sites like serverfault, superuser, etc, the problem is more of a black box and often it's not immediately clear when reading the post what is the cause of the problem. Some answerers need to research it a bit and maybe find a source to link to.

Of course, on-topicness should be the sole criterion to determine on which site to ask a specific question. But it's an intriguing question nonetheless; do you really get quicker answers on Stack Overflow than on Server Fault or Super User? On which site will you get an answer the soonest after posting a question?

1 Answer 1

20

This query compares the average 'time to first answer' across the network (for an adjustable time period; default is the first half of 2018). Answers which are later deleted aren't counted, and I should probably exclude self-answered questions but I doubt that will make much of a difference. Here is a version to play with for a single site.

Tamil Selvan C, who wrote the original question, was right; if you look at the questions posted in the first half of 2018, the average time to first answer on Stack Overflow was 50 hours (just over 2 days), while Server Fault and Super User were really close to each other at 85 hours (3.5 days) each. Users on The Workplace were served fastest; just over 3 hours, followed by Interpersonal Skills (3.8 hours). Community Building was slowest with almost 355 hours.


It's also interesting to see what happens if we try to produce some graphs. If we simply count Stack Overflow questions by the time to the first answer, we get this:

enter image description here
x-axis: time to first answer in minutes, up to 1 day;
y-axis: logarithm of the # of questions, so 6.0 corresponds to 1 million questions

You see that the top of the graph is already at 3 minutes after posting, which I personally think is really fast. (But it does not mean 50% of the questions are answered by that time, not by a long shot.) Other sites show similar but slightly slower 'top' times: on Super User, it's 5 minutes; on MathOverflow, 16 minutes.

Another interesting pattern is this 24 hour cycle; you see that local maxima occur at multiples of 24 hours.

enter image description here
x-axis: time to first answer in hours, up to 1 week;
y-axis: logarithm of the # of questions, so 6.0 corresponds to 1 million questions

10
  • What did Arabic and StartUps wrong so they have to be excluded in that first query?
    – rene Mod
    Commented Jul 28, 2018 at 10:48
  • They gave an average of NULL, making them appear on top of the list. I didn't know why, but now after a run it dawns on me: they simply weren't active during that period. I'll modify the query slightly.
    – Glorfindel Mod
    Commented Jul 28, 2018 at 10:59
  • Ah, yeah, that makes sense I thought you copied the query from somewhere and left the restriction in as an oversight.
    – rene Mod
    Commented Jul 28, 2018 at 11:10
  • 4
    That's all I ever do as a developer: copy, paste and modify :P
    – Glorfindel Mod
    Commented Jul 28, 2018 at 11:12
  • 2
    What could be the reason for the 24-hour period? OPs revising their questions a day or two later at approximately the same hour as when they were initially posted and getting renewed attention? Or some mechanism in the Stack Exchange system? Commented Jul 28, 2018 at 12:33
  • @PeterMortensen excellent thought about the revisions. It's something we can probably analyze further by looking at the revision history.
    – Glorfindel Mod
    Commented Jul 28, 2018 at 12:34
  • That result on Community Building really surprises me. The site doesn't get that many question, but when it does people tend to jump on them, from what I've seen. (Answerers are hungry for questions to answer.) Commented Jul 29, 2018 at 3:02
  • @MonicaCellio that average is over just 7 (answered) questions. After this morning's SEDE refresh, the value has already dropped from 616 to 355...
    – Glorfindel Mod
    Commented Jul 29, 2018 at 10:16
  • @Glorfindel are you time-limiting the questions or something? I assumed it was an average over all (answered) questions on the site. Commented Jul 29, 2018 at 21:46
  • Yes, first half of 2018 but you can adjust the parameters of the query if you want. No SQL knowledge required :)
    – Glorfindel Mod
    Commented Jul 30, 2018 at 5:23

You must log in to answer this question.

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