5

How can we improve our answer acceptance rate?

As I write, there are 18,677 question on MP&T with at least one positive-scored answer, and there are 9,978 questions with accepted answers: a rate of 53%. (Looking at total questions, the acceptance rate is 50%.) I haven't found health metrics for graduated sites, but for beta sites, according to How were the guidelines for a healthy site determined?, 90% is considered healthy. A personal acceptance rate of 70% is suggested in (the accepted answer for) How does accept rate work? Taken together, that would suggest we're subpar.

The "helpful" bar for accepting answers strikes me as both very low and very reasonable. So either we're not very good at providing helpful answers, or we're not very good at accepting them.

Thoughts?

Related: Why is our engagement rate low?


EDIT:

Now that there's been time for initial comments, some reactions.

Accepted answers with negative scores

There are 24 such answers on our site. That amounts to 1 in 400 accepted answers or 1 in 833 total answers. That small ratio seems like something we can live with.

User acceptance vs. community voting

Both provide important metrics. Community voting gives an indication of the overall quality of answers. User acceptance gives an indication of whether individual users are being helped by the answers. A "helpful" rate little better than a coin flip does not suggest a useful q/a site.

Inexperienced users

It's true that most questions come from less experienced users, and the acceptance rate is also lower among those users.1 All the more reason that the more experienced users should set an example both by making a point of accepting helpful answers and also by helping the newest users understand how the site is intended to work.2

The importance of acceptance

To suggest that acceptance is being treated as "so important" is to set up a straw man, and to suggest that reputation gain is an important motivation denies the basic design of SE sites. The point here is only that acceptance of answers has importance.

When searching for information, an accepted answer helps narrow the field. "This question has at least one answer that was useful to the asker." A searcher would be more than justified to prioritize questions with accepted answers over others.

And when looking for questions to answer, the reverse applies. A question for which a helpful answer has been supplied would rightly attract less attention than a question for one which has not.

It's easy to point out pitfalls in both approaches, but this is inherent in all SE sites. Questions may be incorrectly tagged. Wrong or low-effort answers may be upvoted. Titles or keywords might not turn up the needed result. Nevertheless, all have their value, and when well-maintained by the community, they make for an effective system. Responsibly improving our acceptance rate is one way to increase the value of that particular SE feature.


NOTES:

1Users no longer registered account for 3.7% of total questions with an acceptance rate of 46%. Users with Reputation below the Established User privilege account for 75.7% of total questions and an acceptance rate of 48%.

2Among Established Users (Reputation >= 1000), the average reputation is 4877. Users with a Reputation >= 1000 and < 4877 accept answers at a rate of 62.8%. Users with Reputation above 4877 accept answers at a rate of 56.3%.

13
  • 2
    I don't really view the acceptance rate as an issue personally. It's driven by the the user that posts the question not the community as a whole and we get most of our questions from newer users who may not understand the feature. A question with an answer of score of 1 or more is considered "answered" and in that stat we only have 530 questions unanswered (though I've never really like definition of one vote marks a question answered regardless of votes on the question). Users can also accept wrong/inaccurate answers which are then downvoted. I'd rather have us focus on voting in general.
    – Dom Mod
    Commented Sep 24, 2020 at 13:28
  • Here is a query that gives an idea of the accept answers with a negative score on the site: data.stackexchange.com/music/query/1298260/…. This is why I kind of tend to lean on us as a community voting high quality content over accepted answers. They are still an important part of the identify of SE, but I don't feel like have a low accepted answer rate really affects the quality of the site.
    – Dom Mod
    Commented Sep 25, 2020 at 3:46
  • 3
    @Dom We have our share of highly active users that spam low effort answers everywhere. That has to have some impact on the acceptance rate. To find an actual answer you have to read through all the spammers, and since most of these low effort answers are also low quality (and many barely relate to the actual question at all), the askers might not feel that the question is actually being answered. Having to deal with spammers discourages participation from both ends. Commented Sep 25, 2020 at 4:02
  • 2
    @VonHuffman Interesting point. I almost wish it were possible we could see these numbers after filtering out all of the "who cares, do what you think sounds good!" answers ("answers").
    – Richard
    Commented Sep 25, 2020 at 10:01
  • 1
    @VonHuffman that can be fixed with downvoting low quality answers and flagging it if it's really not an attempt as an answer. IMO that problem is orthogonal to this. Someone can accept any answer, not necessarily a high voted one and we have an avergae of over 2.5 answers per question so even in that case they are getting another answer.
    – Dom Mod
    Commented Sep 25, 2020 at 13:41
  • @Richard we can probably repurposed an already made SEDE query for that. One that already scans for a phrase would be ideal.
    – Dom Mod
    Commented Sep 25, 2020 at 13:42
  • 1
    @Richard he's a quick version I did. It's not perfect, but it can be used as a building block: data.stackexchange.com/music/query/265839/…
    – Dom Mod
    Commented Sep 25, 2020 at 13:58
  • @Dom Brilliant! This is a great building block, and certainly one we could use for this.
    – Richard
    Commented Sep 25, 2020 at 14:30
  • What actually constitutes a 'spam answer'? Most wrong answers get dv'd, and there are very few bland answers, so it can't be those, can it?
    – Tim
    Commented Sep 27, 2020 at 15:46
  • 5
    Consider this: "Not every user comes back to accept an answer", those including unregistered users who lost their cookie-based access and cannot accept the answer anymore.
    – Andrew T.
    Commented Sep 28, 2020 at 5:45
  • 1
    One of the problems with MP&T is that a lot of perfectly good questions end up having subjective or opinion-based answers. There are often several answers to a question which are all good but which disagree in whole or part - what is the asker supposed to do? Commented Oct 2, 2020 at 16:31
  • 2
    I suspect if we strictly shut down all opinion-based questions there wouldn't be much of a site left... Commented Oct 2, 2020 at 16:32
  • 1
    The opinion whether an question or answer is opinion based or not - in MPT - seems very opinion based to me! Commented Oct 6, 2020 at 11:50

1 Answer 1

4

Why should we deem acceptance as so important? Often questions are from casual visitors to the site - regard the number of newbies - who maybe don't even ever return to read their answers. If they do, there's no compulsion to accept a particular answer - and they may not understand some anyway. There have been questions I've posed (hardly a newbie!) that still have no accepted answer. I regard it as my right to decide, for whatever reasons, and I always have those!

One purpose of this site is to build a compendium of information appertaining to music (practice and theory), and the questions and their answers do this well. Acceptance may or may not be helpful to anyone. I've seen accepted answers that were wrong. Who does that help?

True, it's nice to have one's answer accepted, but what criteria does that give it? OP likes it best. So what?

2
  • Some very wise person has defined "health metrics", so ... Computer says no. Thou shalt not question health metrics, programmed by fellows with compassion and vision. Commented Oct 3, 2020 at 19:42
  • @piiperiReinstateMonica - that's a great pity. I was hoping the programming would be achieved using logic, and stuff like that.
    – Tim
    Commented Oct 4, 2020 at 6:48

You must log in to answer this question.

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