17

There was some chatter in our mod room about the answer rate, currently at 58%. I've made the argument before that such a low number is natural, and even good, for a site like this:

% Answered

I detest this statistic for this site. In fact, I like it rather low. Yes, if it was 40% I would be worrying. But if it was 90%, I would be suspicious that most of the questions here were too easy - not scoped well enough to be specific.

In my view, the reason this number is low compared to other sites is that we want very specific questions - and sometimes (often!) software simply doesn't exist that fulfills all the requirements people put in their question. This is a good thing, as it means our questions are very specific!

A few things have changed since 90 days into the beta, but I still stand behind the main point there: Our % answered will and should be lower than most sites, due to the quality requirements we have for questions.

That said, it'd still be nice if it was higher. 58% isn't a terribly good number. How do we make it higher? Get experts who know about the software.

I put together this query to understand where the problem lies. It gets all the tags with >50 questions, and calculated what percent is unanswered*:

enter image description here

I noticed something right away here - the top offenders seem to be software development type tags. I downloaded the data and put it in a spreadsheet, classifying each tag as either "Stack Overflow-type" or "Super User-type".

Notably, for SO-type tags, our unanswered rate is 42.71%. Compare that to the rate for SU-type tags, 33.6%.

Overall, this is the lowest rate on the network. As I said, I think at least part of this is inherent to the type of site we're running and the quality we require of questions. Still, we could do with more experts.

The difference between the unanswered rate for SO-type questions and SU-type questions seems to be significant. So, my question to you:

Is this number anything to worry about? If so, how can we better reach experts, especially in the software development field?


* Note that my definition of 'unanswered' here is "doesn't have any answers", which is different than the one used to calculate the Area 51 statistic. Also, this is counting closed questions.

| |
  • 2
    Good points (and great idea for the stats)! Just your table has a calc error in its last column – or our answer rate would be higher than 99% for those tags :) Try ROUND(CAST(COUNT(*) as float)*100/CAST(total_count.total_count as float),1) as percent_unanswered in line 4 of your query. – Izzy Sep 4 '15 at 22:35
  • @Izzy I'm confused - doesn't that just change the format of the percentage into a more human-readable version? I'm not seeing any significant differences between the two ;) – Undo Sep 5 '15 at 1:16
  • 3
    Undo, that's only part of it. See the *100, that's the important part: e.g. for c#, it's not 0.58% unanswered, but 58% :) – Izzy Sep 5 '15 at 9:12
  • 1
    @Izzy Ah, right. I'm just used to reading them as x/1, not x/100. Sorry :) – Undo Sep 5 '15 at 14:52
  • 1
    Yeah, thought so. But then you can't call that "per cent" (remember, a "centurio" is responsible for 100 :) – Izzy Sep 5 '15 at 14:56
  • Do I get that right: We want to be that "percent_unanswered" to be 0 (or as low as possible), right? That would mean that all questions have answers. – Angelo Fuchs Sep 5 '15 at 15:41
  • Right, @AngeloF – Undo Sep 5 '15 at 16:00
  • Maybe those big tags should be split, to reach people active only in some subtag thereof? – Nemo Oct 6 '15 at 12:09
  • Your "percent_unanswered" isn't percent unanswered, it's proportion unanswered. The factor of 100 makes a difference. Which is sort of what Izzy was saying. As it stands, that column strongly mis-represents the situation, and I know that wasn't your intention at all. (and the number of decimal places is nonsensical. Two significant figures would be sufficient, more is noise) – 410 gone Oct 12 '15 at 11:11
8

One thing I've noticed is that many questions that should be asked on Software Recommendations SE are instead asked on the Superuser SE.

For many of these questions, no one flags them to indicate that they belong on the Software Recommendations SE.

Furthermore, when I have flagged questions on other SE sites, the Software Recommendations SE is missing from the list of possible other SE sites that the SE UI presents. Fixing this bug would likely help.

More importantly, educating the folks on Superuser about Software Recommendations would be of value.

| |
  • 1
    I've already asked in a few SEs to redirect to softwarerecs, but some complain softwarerecs is beta, e.g. meta.webapps.stackexchange.com/q/2636/18147... – Franck Dernoncourt Sep 4 '15 at 3:30
  • 2
    It's a policy on SE to (almost) never migrate to beta sites. Still, mods can do so – and I agree they should, but only for high quality questions (as there's the other rule: don't migrate crap :) plus, if they're not familiar with our "standards", after asking in our chat room. // Btw: AFAIK that "universe list" doesn't include betas, does it? I didn't do a thorough check on that, but I'm pretty sure about it. – Izzy Sep 4 '15 at 22:41
  • 2
    The migration limitation should be gone now that SR is out of beta – Thomas Weller Sep 10 '15 at 14:07
7

I think it lies (partially) in the low votes / work ratio. (Please note that I don't have direct data for that, just my personal observation)

So, I see part of the problem not in getting people here, but in keeping them here.

Good answers on SR.SE tend to be longer and take more work then good answers on sites like SO. But on the other hand, the quality requirements and personal time here cause people to be less likely to upvote stuff.

So, longer posts take longer to read, fewer people read it and those who read it are less likely to upvote them.

So, if we consider reputation as the primary reward system AND we require long answers, then we must upvote them more.


Notes:

  1. Long answers don't get rewarded on SO either, but one can make good short answers there, but not here.
  2. I think our quality requirements are good the way they are, we must not compromise on them.
  3. I do not know of a way to make longer answers more visible.
| |
7

IMHO this has two reasons:

  1. software developers (Stack Overflow) and system administrators (SuperUser) might have very special needs - and the required program simply does not exist
  2. there is no migration path (yet). We should make sure that SR appears here on SO:

    StackOverflow migration paths

    I personally would flag questions like that, which in turn would train SO users. I didn't ask for moderator migration until now, instead vote for off-topic. This should no longer be needed since SR is out of beta.

    This will not immediately reduce the number of open questions, in fact we might get a lot of crappy new questions here. But over time, Stack Overflow and SuperUser users will find this place useful, ask better questions and answer other's questions.

| |
  • 1
    We're graduating! There's gonna be one now! :D – ᔕᖺᘎᕊ Sep 10 '15 at 19:16

You must log in to answer this question.

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