I am sometimes surprised at certain questions that have received huge numbers of views. For example, I just stumbled across this question. I mean it no disrespect - it's a perfectly nice question - but it only has three votes and is quite esoteric and closed as homework-like. I can't imagine that people often search for keywords that lead to this question, or click on it based on its title. Nevertheless, it has almost 17,000 views, which almost puts it within the highest percentile of the most widely-viewed questions on the entire community.

Could this be the result of bots or a bug in the analytics? Or is there something about questions like this that I'm missing which makes them immensely popular? (Again, I mean no disrespect to the OP and I don't mean to be snarky - I'm genuinely curious.)

  • 3
    " it has almost 17,000 views, which I believe puts it well within the highest percentile of the most widely-viewed questions on the entire network." Not even close. That's pretty active for Physics.SE, but the really high activity sites generate multiple 10,000 view questions per day. See stackoverflow.com/help/badges/28/famous-question for evidence. – dmckee Aug 11 at 1:24
  • 1
    @dmckee Sorry, I edited to say entire "community", not "network". I was basing that on the fact that only 2490 out of 123839 physics questions have the "Famous Question" badge, and 17000 is significantly past the threshold, so assuming a rapid falloff in the distribution, this question's view count should be above the median Famous Question's. – tparker Aug 11 at 3:02
  • 1
    I usually protect low-level questions that have many views. The recently protected list should give you many examples of questions with thousands of views and few upvotes (e.g., physics.stackexchange.com/q/301068/84967, at 11k and +1/-1 votes). – AccidentalFourierTransform Aug 11 at 3:09
  • @AccidentalFourierTransform Yes, these are exactly the kinds of questions I'm talking about. How do you find these questions? Do you have a way of knowing if these views came steadily or in a sudden surge? – tparker Aug 11 at 3:39
  • @tparker I mostly stumble upon them in the Late Answer queue, or when Community bumps them into the front page. And I have no idea where the views come from, but I always assumed it's due to google. SE is usually positioned to the top of a google search, so I guess that's where most clicks come from. – AccidentalFourierTransform Aug 11 at 3:43
  • 1
    @AccidentalFourierTransform @ tparker If finding questions with low score and high viewcounts is a goal in itself, it is easily doable using SEDE. – Emilio Pisanty Aug 11 at 11:42
  • 1
    Also: from some initial testing on SEDE, this just misses out on the 99th percentile by views, which at about 18,300 views. – Emilio Pisanty Aug 11 at 11:57
  • @AccidentalFourierTransform I see you took that 'goal in itself' bit to heart ;-) - I'm glad to see this one got spared, at least. I've updated the query to include a views/mo field which might also be useful. – Emilio Pisanty Aug 11 at 18:18
  • @EmilioPisanty I cannot protect that one, it doesn't have any deleted/downvoted answer. I would if I could :-P – AccidentalFourierTransform Aug 11 at 18:36
  • Part of the answer may be psychological: When a particular question begins to get a large number of views, others may wonder what the fuss is about and view it, increasing the number of views ….? – jim Aug 20 at 18:48
up vote 9 down vote accepted

Unfortunately, search analytics are kinda crap, and we appear to have lost most of the historical data for Stack Exchange sites in Google Analytics. So I can't really give you much in the way of inside info here, other than to note that the view counter doesn't rely on any external system - more info here. So while view counts aren't perfectly accurate, they're probably pretty solid over long periods of time (which is the case here).

So, let's do a bit of amateur detective work via archive.org:

And today at 64 months it has 16804 views, or about 366 views/mo since the last capture on Archive.org.

This jibes pretty well with knzhou's theory: regular crops of students googling for info and landing on that particular question. So I'd probably go with that as my assumption too.

To confirm, I sampled the server request logs for a few days over the past year, and all hits to that question came from... Google.

  • Thanks for the great detective work. Looks like you've got it, although I admit that I'm a bit mystified as to exactly what terms tens of thousands of people could be Googling to get to that particular question. Maybe we should be concerned that so many high school and undergraduate student are apparently studying polonium in such depth... – tparker Aug 11 at 19:27
  • 1
    @tparker That question is top ranked under 'alpha particle kinetic energy' on my machine - I'd wager that that's a substantial contributor to that traffic. – Emilio Pisanty Aug 12 at 11:18
  • 1
    @tparker Because high schoolers have to calculate the kinetic energy of an alpha particle all the time. Some particular calculations gain almost mythical important in high school science because they can be performed with nothing more than a calculator, arithmetic knowledge, and a lookup table; these questions are easy to mass-produce and easy to grade. Literally 25% of the content and over 25% of the points of a unit on "nuclear physics" in a "modern physics" high school course will probably be devoted to this one calculation alone. – knzhou Aug 12 at 11:23
  • If one big school system uses a textbook with this question in it, that would account for many times this number of views over time. (The system I taught in had >140k students, and they were ALL required to take Physics when they get to 11th grade -- if they didn't drop out first.) – Jeffiekins Aug 17 at 15:54
  • I think the historical view count of the questions could be an useful data for tuning the sites for growth. You have the logs, maybe you could import it into some local db for data mining (if you don't do that already). – peterh Aug 17 at 21:34

It's not the HNQ effect; the upvote/view ratio is all wrong for that, and the views come in over too long a time period. It's just the fact that StackExchange is good at search engine optimization, and the market for high school homework help is huge.

The unusually popular questions have one of two features. Either they contain the exact text of a commonly assigned high school textbook problem, or they contain a common phrase, so that somebody googling "how to convert enthalpy from J/g to kJ/mol" will land on the question with title "Converting enthalpy from J/g to kJ/mol". If a question's title or body matches what the typical high school student is searching for a little bit better, it'll get 100 times the views.

Speaking as a former high school student, I'm sure you'll all agree high school science is terrible. Most things are explained poorly, and many homework questions hinge on arbitrary conventions and unstated assumptions, when they're even correct to begin with. (Last year I helped my sister through a chemistry class where she kept getting zeroes for writing "8.392 kJ/mol" instead of "8.393 kJ/mol". Apparently, not only the way you round but the rounding convention for intermediate calculations must match for credit.)

So it's not surprising that every day, millions of students return home and start googling for homework help. This market is what gives sites like Quora and Yahoo Answers view counts that are ~100x ours, but also their much lower quality. This is the real reason the homework policy exists.

  • How can one see the timing of the view count progression? – tparker Aug 11 at 12:35
  • 1
    @tparker It's extremely unlikely that you'll be able to find that data - as far as I understand, via this answer and this one (also this one), SE doesn't keep historical logs of viewcounts. – Emilio Pisanty Aug 11 at 12:37
  • 1
    Also, please convey to your sister my sincere sympathy and reassure her that science eventually does get better once you clear the initial layer of morons. That story's appalling. – tparker Aug 11 at 12:38
  • @EmilioPisanty Yeah, that's why I'm confused how knzhou knows that "the views come in over too long a period." – tparker Aug 11 at 12:40
  • This is certainly supported by the fact that many of these questions are closed. – Norbert Schuch Aug 11 at 12:50
  • 2
    @tparker Because an question with less than 2 upvotes will not stay on the HNQ list for long, and even questions that stay there for days are lucky to get ~10k views typically; the traffic simply isn't enough to get the numbers we're seeing. Also, many of the questions are quickly closed, so they can't get HNQ views. Another feature these questions have in common is that they tend to be several years old. – knzhou Aug 11 at 12:53
  • @tparker I basically told my sister chemistry was just silly like that, since my chemistry class was just the same way — all bookkeeping and arbitrary unjustified rules. But it would kill me if her physics class next year was just as bad! – knzhou Aug 11 at 14:57

One common answer is the Hot Network Questions sidebar that appears across the Stack Exchange network. Questions that get in the HNQ often attract views (and votes, and don't get me started on that) far out of proportion to the day to day expectations of the site.

As far as I know, being selected to appear in the HNQ sidebar does not leave a audit trail that is visible even to site moderators, so I can't say if that question did or did not partake of the HNQ effect.

  • Isn't there a bit of a chicken-and-egg issue though, because don't questions only become HNQs once they've already received a lot of views? – tparker Aug 11 at 3:41
  • 1
    Well, it doesn't seem to take very many clicks or votes if they happen very shortly after the question is posted. And the list can be a bit of a self-fulfilling prophecy. – dmckee Aug 11 at 4:04
  • More on the audit trail for HNQs on the mother meta: one, two, and a third-party tool here. More on the chicken-and-egg problem here. – Emilio Pisanty Aug 11 at 11:12
  • 1
    But wouldn't you assume the HNQ questions attract significantly ore upvotes? – Norbert Schuch Aug 11 at 12:40

You must log in to answer this question.

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