The cleanup sprints have been great at improving our answer rate but does our answer rate really matter? I suspect "cleanup" brings negative behaviours:
- Closing questions that are hard, unanswered but valid (and otherwise fine)
- Upvoting crappy answers to count the question as answered
- Comment-carpet-bombing "Nobody has done anything in here for a while so I'm going to flag it for removal"
- And oh my god, you've not seen flags until you've been a moderator in cleanup week. People go bananas and it makes it harder to deal with current issues.
All this so we can push off old questions that haven't been answered. Don't get me wrong, that's exactly what it was designed for; to give us a fresh start with a sustainable answer rate but when we do bad things to handle a past inability to answer, we're just compounding original failure.
I'm not saying anything against a future cleanup to help us cope with these numbers but we need to acknowledge that our "answer rate" is not the same thing as "the rate that we are answering new questions" and that's the problem. That's what we need to nail. I think we need to look at how we allow answers onto the site so that six months after the next cleanup we're not back in the very same position we are today. This is exactly the stuff you're talking about.
Hardware problems are killing us; can we improve the process for asking a hardware question?
I think our main issue is there's not a good system for hardware issues (which make up 50% of our questions - pulling that out of thin air but it feels right). I would suggest a wizard (that can be skipped) that poses the following questions, one at a time:
- Are you having problems with graphics, networking or audio hardware? (user selects networking)
- Please open a terminal (link to help) run
lspci -nn | grep netand input that into the following box.
- Is your question in the following: ... (We do a mega-search based on the module, driver, brand. Most users should exit at this stage.)
- Ask your question (we pre-populate the question with the lspci details)
Obviously the wizard would be different for different hardware (graphics might be quite a long process, for example) but that's it. Just a framework for making sure users have searched for the right things and that if they have, we have the right data from the very beginning to answer the problem.
I have talked about implementing something like this in the past but have (obviously) never got around to. I honestly think a more informed search wizard, is going to skim out at least 30% of new questions and make the remaining much more answerable and easy to dupe out.