The Fastest Gun problem (here and here) sort of discourages thoughtful answers on SO and SE. But I see this problem as caused by another one: the habit of many users to ignore upvoting at all.

Here's a simplified process that most questions go through:

  1. A user asks a question
  2. Several permanently active users give answers within a day
  3. Answers are upvoted and one is accepted
  4. No upvotes for answers that arrive later

Step 3 provides minimal answers. Step 4 is a problem because the question gets regular traffic at the rate about .4 views per day (source), while the quality of answers stagnates. How it happens?

People reach answers mostly from search engines:

enter image description here


We can't expect from people looking for an answer to provide one. But these visitors don't upvote as well:

enter image description here

(source, see for more of good insights)

In brief, active users don't reach questions in retrospect and passive users (searching for answers) don't leave their feedback.

My point is, this feedback, if present, would encourage late thoughtful answers and it'd be useful to understand why people coming after answers don't upvote them.

Perhaps, you can share your ideas why late visitors don't vote and how to fix it. Personally, I'm out of data on the previous attempts to solve it and closed data on user behavior (their in-browser actions before closing the page with the answer they've found).

  • 2
    I think I read that around 90% of the traffic comes from Google. I don't know what portion of this people have an account, so that they can up/downvote. Also, I see a funny trend here: if a question/answer got up/downvoted in the first days, then it is going to get a lot more up/downvotes on the long turn. I have answers in popular questions in which I get few upvotes every week, whereas in others not. It kind of looks like we have a tendency to "emulate" what the people before us did. So: let's vote to make future visitors emulate it! – fedorqui 'SO stop harming' May 12 '15 at 22:01
  • Do your sources count all visitors, only visitors with an account, or only visitors with an account and the neccesary rep to vote? The number of votes may well be proportional to the later number of visitors, but I suspect you count the former. – HugoRune May 12 '15 at 22:03
  • @fedorqui Yes, I also couldn't find a way to separate out registered users. Though I believe registration is not a huge obstacle if a person wants to vote, which we want him to do. But legitimate users don't vote as well compared to their activity just after the question was added. – Anton Tarasenko May 12 '15 at 22:19
  • @HugoRune That's SimilarWeb's stats. Added the link. – Anton Tarasenko May 12 '15 at 22:20
  • @Anton: registration isn't sufficient to vote. You need 15 rep. – Mat May 13 '15 at 5:18
  • @Mat There're 330K eligible voters (> 15 rep) who never voted: data.stackexchange.com/stackoverflow/query/313777/… – Anton Tarasenko May 13 '15 at 9:30
  • 4
    I don't see anything in your question to support "late users don't vote" Is it because traffic comes from search engines? I'm an active user and often arrive at questions from search engines when I'm solving a specific problem for work. I upvote all the helpful answers, and the question for asking in advance so that I got answers right away. I don't think I'm alone in that. – Kate Gregory May 13 '15 at 12:26
  • @KateGregory Added the stats on that. – Anton Tarasenko May 13 '15 at 13:11
  • The point you make is important, but the correlation is a bit fragile: while active users mostly vote soon and the large majority of the visitors is passive, many late visitors may be active ones. It's also possible that in certain situations the votes given in the long run matter more than the early votes. Like, is there a difference depending on the score of the question or the answers? on the presence of an (accepted) answer? Do hasty answers discourage answers in the long run, compared to questions without answers (which need more attention? – Nemo May 13 '15 at 14:37
  • @Nemo It's not about active users who vote (they're a tiny minority), but about passive users who don't vote. They use answers but don't incentivize (signal) authors to write more or revise. – Anton Tarasenko May 13 '15 at 15:35
  • @Anton, the two are linked. Of the total amount of votes, those coming from less active users are the complement of those coming from very active users. You have to decide which of the two populations to cater for (you seem to want to get new voters) and to see what they're doing currently. But everything is mixed in the graphs above. – Nemo May 13 '15 at 17:23

I'll give it a try, since no answers have been posted so far.

The problem: Improve popular questions and answers over time

Subproblem 1: Finding questions and answers worth improving

Popular questions are those that get most traffic. Let's say, the threshold is 50% of the total views. These are few questions because the distribution is highly skewed:

enter image description here

(distribution of questions by the number of views, source)

This formulation of the problem is preferable to asking to write good answers right away because it's not always evident that the question is hot, so it's better to save efforts until it becomes evident.

The hotness becomes evident when questions continue receiving views over time.

But users can't even track the questions that get hundreds of views each month over years. The "Hottest Questions Today" section covers only short-term peaks in activity (generally, the SE interface encourages the fastest gun behavior, even though these daily hits soon get forgotten). The dynamics of views is unavailable on data.stackexchange.com either. It has "total views", but not "views on week/month/year t".

Subproblem 2: Late answers and improvements are not encouraged

The reason is that late voting (week plus after the question was posted) is rare, otherwise the votes wouldn't have exponential distribution on a logarithmic scale:

enter image description here

Users are roughly divided into two groups: (1) active community who browse the front page, give answers, vote, and moderate; (2) passive readers who google questions, read, and close the page—the bounce rate is 52%. The active community monitor changes. But since changes are rare, late voting by community members is rare.

I found no mechanism that encourages late revisions or at least indicates their necessity.

Solution: Encourage passive users to vote

Passive users come from search engines and they don't vote, as seen from the data in the question above.

Putting the problem of registration aside, there're 330K of users who can vote but never did:

enter image description here


How to encourage these eligible users to vote?

I suggest two ways:

  • Reminders while the user is looking for answers
  • Delayed reminders, when the user has time to vote

I can think of traditional reminders, such as:

  • Popups (done by SE devs before to encourage upvoting for questions)
  • Emails (SE does send notification emails)
  • Special pages for listings (SE has a page with "hot daily questions," for example).

All these forms are practiced already.

It's harder to make such reminders friendly. The best way is to reduce the reminders to the necessary minimum, so that the reminders concern only:

  • Questions with daily traffic above average (quality and, hence, feedback are important)
  • Answers that captured the user's attention.
  • As indicated by mouse interactions (clicks, selections, hovers) with specific answers. This says which answer the user had taken but didn't vote for.
  • Limited number of questions per time period.

Asking for three votes per week by email would yield millions of feedbacks on important questions.

These upvotes, in turn, encourage authors to improve existing and write late answers.

  • I find this answer contradictory... if active users only look at recent/updated questions, why would they care about score of old questions? – Nemo May 13 '15 at 17:25
  • @Nemo Because right now active users receive only weak signals about the quality of old answers. These signals come mostly from other active users and the signals are weak because active users are few. – Anton Tarasenko May 13 '15 at 17:53
  • 1
    ok, but how do those signals work? Are higher score questions "pushed" more to the active users, e.g. in the "answer another question" boxes? If someone has time, it would be great to check the data to see if there's higher late activity after question upvotes. For sure upvotes don't bump questions even though it was proposed. A mechanism I do understand is that (deletion by) downvoting and copyediting helps keeping the backlog cleaner. – Nemo May 14 '15 at 5:39
  • @Nemo There's no channel between question upvotes and active users for old questions with answers. I don't expect any correlation here until SE creates such a channel. I do expect correlation between answer upvotes and ex-post edits (improvements), but unable to show this because I didn't find the way how to get the history of edits from data.SE. LastEditDate is insufficient for this. – Anton Tarasenko May 14 '15 at 13:46
  • I often see a banner when visiting old SO questions from Google that says if I found the question/answers helpful, I should consider upvoting them. Doesn't this prompt already do what you want? – Troyen May 18 '15 at 23:14
  • @Troyen The banner that appears after auto login? I don't believe it works well for this purpose. – Anton Tarasenko May 19 '15 at 0:15

You must log in to answer this question.

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