Optimum posting time?

March 9, 2009 11:04 AM

What are the peak hours of MeFi and AskMe?

What's the best time (meaning both time of day and day of the week) to post to this site, if you want your post to get tons of traffic and responses?

I'm asking because my last two AskMe questions didn't get as many answers as I hoped they might. Both concern phenomena that most users probably have experience with and hypothetically could have weighed in on, but both posts seem to have rolled to a stop with under 20 responses. The questions were read by a lot of people as they were both favourited quite a few times, but I wonder if I'd have gotten more responses if I'd posted at a different time in the week instead of *cringe* Saturday and Sunday evenings.

Thanks!

posted by pseudostrabismus to MetaFilter-related at 11:04 AM (50 comments total)

There's a graph by cortex here that might help.
posted by Artw at 11:15 AM on March 9

This gets asked a lot.
posted by vacapinta at 11:17 AM on March 9

Man, that graph really is oblique.
posted by cortex at 11:22 AM on March 9

Shouldn't the optimus prime tags be a single tag?
posted by cjorgensen at 11:28 AM on March 9 [5 favorites]

Clearly, according to the graph, one should strive to post within the bluish area.
posted by Mister_A at 11:45 AM on March 9 [2 favorites]

I've always operated under the assumption that posting rate and reading rate are proportional. So more people read the site on weekdays, but more people post too, so your question falls off the front page more quickly. So if your question has global appeal, I doubt it matters. I would love to see real data to confirm or deny this, however.

The problem is that not every question has global appeal. The best example is locality-specific questions. I remember one person who asked a question about the UK at the British bedtime. Everyone who could answer was asleep, and by the time they woke up it had fallen off the front page of AskMe, pushed off by lots of American questions.

The best thing to do is to figure out when the For example, I waited a couple of days to post today's question about American retirement plan stuff during the American weekday morning because I thought that a proportionally greater number of people who would have an answer would be checking MetaFilter at that time.
posted by grouse at 11:49 AM on March 9 [1 favorite]

The questions were read by a lot of people as they were both favourited quite a few times,

Is there any data on the number of favorites vs the number of responses? I'm guessing there's not real relationship. Aren't there a lot more lurkers than active participants?
posted by Brandon Blatcher at 11:58 AM on March 9

I've found the best time to post is when I remember what I wanted to ask, and the worst is when I have forgotten.
posted by Astro Zombie at 12:09 PM on March 9 [1 favorite]

I am completely and utterly shocked by this graph. That is the average number of comments to AskMe questions posted at that time. It is startlingly flat. The conclusion: It doesn't matter what time you ask your question. What matters is how you ask it and whether anyone has the answer.

The infodump is dead (for now), so I had to use the last file I had, which was from last October.

Here's the script I ran:

perl -lane 'next if (/^\D/);
  my $t=substr($F[3], 0, 2).substr($F[3], 3, 1); $n{$t} += $F[4]; $p{$t}++; }
  for my $t (sort keys %n) { printf("%.1f,", $n{$t}/$p{$t}); ' postdata_askme.txt

posted by Plutor at 12:11 PM on March 9 [5 favorites]

Twenty responses seems pretty reasonable. In fact, in one of your linked questions you got "so many awesome responses".
posted by DU at 12:14 PM on March 9 [1 favorite]

Yeah, at least you got a lot of responses. There are questions floating out there with nothing a'tall. My general feeling is question formation plays a part, but you're better off during business hours on weekdays. Worst time seems to be friday from five to ten p.m. I think there are so many questions now, that you probably get about an hour of focus if you're lucky. If it's a problem to be solved, as opposed to "help me with a list of songs like x, name my x," then after that hour, you've drifted past the core systems and you might as well just enjoy your week of hypersleep and after the 57 years of internet time, you can try again.
posted by cashman at 12:31 PM on March 9

Ah, you misunderstand me, DU. Perhaps on purpose.
I'm not complaining that there was a lack of awesomeness within the 17 answers to that question- a high percentage of those answers were indeed awesome, so my note was a genuine attempt to give due praise to the users who replied awesomely. My question is, why didn't more users reply?

That question (essentially, "help me stay focussed during self-directed work on my own projects") is a broad call for suggestions on a subject which should be an area of some expertise for every single user. 79 people favourited it, and probably hundreds of people read it. So to my eye, getting only 17 responses on a question that open-ended seems really low. 17 strong responses, for sure, but still. I find it hard to believe that less than 25% of the people who liked that question had an answer to contribute to it, and I wondered if the time of the day or week was to blame. Like, maybe, particularly for that specific question, the best time to post would have been a weekday afternoon slump in the US, so all the procrastinators who were in the very process of being unfocussed at work could use that time to give me the advice they themselves were ignoring? So very meta.

Anyway, Plutor's answer- it's kind of the same no matter when you post- surprises me, but I'll operate under that assumption and maybe work on my phrasing to lure in more responses.
Thanks!
posted by pseudostrabismus at 1:01 PM on March 9

I was wondering the same thing pseudostrabismus. I was so sure my third FPP got less traffic than I expected because I posted it at 4:something in the morning.
posted by SheMulp AKA Plus 1 at 1:04 PM on March 9

I believe awesomeness is also a factor. If I see awesomeness in the answer, I often do not chime in. Why should I? Someone already answered it awesomely.
posted by typewriter at 1:14 PM on March 9

I find it hard to believe that less than 25% of the people who liked that question had an answer to contribute to it

I don't understand why you're equating favorites with answers.
posted by Brandon Blatcher at 1:15 PM on March 9 [1 favorite]

You should be grateful. There are precisely 16 general plans to stay focused, and they are conveniently outlined for you with no noise. Would that all of our AskMes could go so swimmingly.
posted by owtytrof at 1:17 PM on March 9

I find it hard to believe that less than 25% of the people who liked that question had an answer to contribute to it, and I wondered if the time of the day or week was to blame.

There's a few interesting things that jump out at me about this (and I mean this in an "I find people's metrics fascinating" sort of way, so forgive me for picking it apart):

1. Why would having an answer to a question be the primary motivation for someone to favorite that question?

People can obviously use favorites for whatever they like, but it's a safe bet that a lot of the favorites that are assigned to AskMe posts are genuine bookmarks of some sort or another. Bookmarking something because you're curious about the answer or the resolution is pretty straightforward, but bookmarking something specifically because you answered it seems less so to me.

2. The number of people who had an answer to the post is less than 25% of the number of people who favorited it, and the number of people who answered the post and favorited it is smaller yet:

- 80 people (as of right now) have favorited your question.
- 12 different users (yourself excluded) answered the question with one or more comments.
- Of those 12, only three (maudlin, citron, and Wolfster) actually favorited the question.

So only about 4% of the people who liked that question (for "liked" = "favorited") actually had an answer to contribute to it. The rest of the favorites posted no answers, and the rest of the answerers didn't favorite it (thought that's not to say they did or didn't like it).

I'd turn it around and say the real question here is, why did the question get so many favorites? 80 is a lot. It's well above average, by an order of magnitude by my estimate.

My answer: because it's asking about something a lot of people deal with or worry about. It's what I've come to think of as General Reference Question, a question whose answers are likely to make an interesting or educational resource for a wide swath of people. And so as a given mefite scans the AskMe index, they see the question, click through, read the more inside and some of the answers so far, and think, "hey, I should get back to this", and they favorite it for future reference.

And then, ironically, they are distracted by a dog that has a puffy tail.

I think there are some interesting drill-down possibilities for more detailed analysis, here. One thing that Plutor's (boggling, worldview-shattering) graph does not capture is weekly cycles—are there more answers on Monday than on a Saturday, or specifically for a question asked at 10 am on Monday vs. 10 am on Saturday? All we get from this graph is that the overall distribution for aggregate 24 cycles is even, without any indication of e.g. larger-periodic variations.

And is there a relationship between number of favorites and number of answers? Variation in either by category (something I should add to the Infodump at some point)? What's the average percentage of answerers who also favorite, and favoriters who also answer?
posted by cortex at 1:30 PM on March 9 [2 favorites]

79 people favourited it, and probably hundreds of people read it. So to my eye, getting only 17 responses on a question that open-ended seems really low. 17 strong responses, for sure, but still. I find it hard to believe that less than 25% of the people who liked that question had an answer to contribute to it, and I wondered if the time of the day or week was to blame.

As one of the people who favourited the attention span question (because I share the problem and wanted to see if there were any good ideas in that thread), who attempted an answer (it's not a full solution, but it has helped me) and who favourited some of the awesome answers and awesome related threads, I think what you're seeing is a completely reasonable discrepancy between interest in a topic and the ability and/or confidence to give a good answer. This was a hard question. It was going to get relatively few answers compared to the typical cat misbehaviour question, which may also get a lot of favourites, but as you said yourself, the few people who respond are more likely to give thoughtful, hard-earned advice. And as typewriter said, the more awesome the answers, the less likely it is for other people to jump in. If I'd come across the question a couple of hours later, I would have been much less likely to submit my answer because other answers covered similar advice but in a more comprehensive way.

Now, the cover letter question hit another MeFi problem: a fair number of responders gave exactly the type of advice you explicitly said that you didn't want, and you had to gently shunt people back on to the path before you got a couple of good examples. I don't think changing the time of day would have changed that situation much. The people who may have had that kind of information most easily at hand -- HR recruioters , mostly -- may form a pretty tiny part of the AskMe readership, too.
posted by maudlin at 1:30 PM on March 9

cortex: One thing that Plutor's (boggling, worldview-shattering) graph does not capture is weekly cycles—are there more answers on Monday than on a Saturday, or specifically for a question asked at 10 am on Monday vs. 10 am on Saturday?

Yeah, I was taken aback, but once I've thought about my own browsing habits it makes some sense. When I keep up with AskMe I read it through rss feeds (with MeTa and the Blue I visit the actual site) so it doesn't actually matter when questions are asked, I read everything I'm likely to have anything to say about. At least for this AskMe browser the time of the day when questions are posted is immaterial as to whether I'll answer it.
posted by Kattullus at 1:44 PM on March 9

Jaisus, cortex, don't you have a job?


Oh, right.
posted by Mister_A at 1:49 PM on March 9 [2 favorites]

Kattullus, yeah, that's a good point. Timeshifting via RSS is probably a pretty good equalizer. As much as anything, I'm just surprised there isn't even a moderate dip in the wee hours.

It'd be interesting to look at a graph of when answers arrive, though, independent of the threads they arrive in, and to look at the like overall rate curve of answers in threads—sure, you got 20 answers, but what did the gaps between those answers look like, and how does that depend on when you posted the question, etc.

Oh, right.

I lead a charmed life.
posted by cortex at 1:56 PM on March 9

Are you looking for a million ideas, or just a few that will work. Maybe the ones that were posted already said what someone else wanted to say, so there's nothing to add but "ditto," which doesn't help your situation.

The number of favorites and comments shouldn't matter. Favorites are nice, but it just means that someone wants to know more on the topic.
posted by filthy light thief at 2:08 PM on March 9

Oops, yeah, Blatcher & cortex, I phrased my point wrong- I didn't mean to suggest that all the answerers had also been favouriters, as it's clear that isn't the case. I was trying to compare the relative numbers and I phrased the comparison awkwardly.

But. Now that you mention it, you know, I AM also surprised that more favouriters weren't also answerers, because if you care about how to stay focussed, you probably have a few tricks on staying focussed. There's not one right answer, and every one of us should have a useful tip or three. Ironically perhaps, as cotex points out, some of us are unfocussed enough that we get distracted by canine poof before answering the question, but still.

What I'm concluding now, after reading typewriter's & maudlin's points, is that maybe there are simply not as many "stay-on-task" hacks as I'd hoped there might be. Maybe the answer really is simple enough to be comprehensively summed up in 16 posts, and maybe that means I should get the hell out of this thread & back to work (sob).
posted by pseudostrabismus at 2:12 PM on March 9

With the attention span question, I think part of it was bad timing that had nothing to do witht he time of the day/day of the week. The question posted just after was very similar. I actually remember deciding that I only had time to answer one of them; it's likely that similarly, people responding just picked one, and that normally, either post would have gotten at least the comments from the other post if it had been on ask's front page alone.
posted by PhoBWanKenobi at 2:17 PM on March 9 [1 favorite]

Plutor - cortex posted a mirror of the January infodump. It doesn't change much - it ranges from 12.0 to 15.4 per question.

Scattergraphing comments vs favorites intrigues me. And I should post the script for my username analysis.
posted by Pronoiac at 2:21 PM on March 9

Also, your cover letter post was really specific in that you were asking to see examples. I didn't feel comfortable posting mine online and didn't feel like going through the effort emailing you one (no offense!). But I would have been much more likely to just share tips, or good websites with examples. I think asking people to share personal information, particularly personal, work-related information, is always going to get less of a response than a call for general advice.
posted by PhoBWanKenobi at 2:22 PM on March 9 [1 favorite]

Yeah, I suspect that the weekly variation is much more important than the hourly variation.
posted by grouse at 2:29 PM on March 9

I didn't feel comfortable posting mine online and didn't feel like going through the effort emailing you one (no offense!).

Yeah, I have a really good cover letter, no way I'm sharing it.
posted by Brandon Blatcher at 2:30 PM on March 9 [1 favorite]

That question (essentially, "help me stay focussed during self-directed work on my own projects") is a broad call for suggestions on a subject which should be an area of some expertise for every single user. 79 people favourited it, and probably hundreds of people read it. So to my eye, getting only 17 responses on a question that open-ended seems really low.

The focus groupy nature of this reply, coupled with the "help me SEO-ify AskMe" nature of the question is kind of icking me out. But to answer your actual question: You have an obvious chatbait question full of buzzwords like "productivity" that nonetheless goes on tl;dr.

MeMail for an address to mail my consulting fee to.
posted by DU at 2:38 PM on March 9

Hmmm, looking at your past comment history, you might have been expecting a lot of answers based on previous popular questions of yours. But those questions were more universal and conversational, which helped bring in answers.
posted by Brandon Blatcher at 2:50 PM on March 9 [1 favorite]

I am completely and utterly shocked by this graph.

Further proof of the Efficient-AskMe Hypothesis (EAH), which states that it is impossible to consistently outperform the average MeFi (i.e. get more responses) by using any information that MeFis already know (the times when most people are online). The EAH also hypothesizes that it's equally distributed over the week, or else it will become equal once someone posts one of these graphs.
posted by Durin's Bane at 4:16 PM on March 9 [1 favorite]

I think the cover letter question is kinda subjective. I've submitted and reviewed tons of cover letters and would be thoroughly annoyed by the Craigslist example. Plus, if someone wanted to interview me based on a letter like that they would be really surprised that my personality doesn't match the bounding out of their seat puppy-dog optimism the letter shows. I have experience in my field and my cover letters are brief, well written, and point to my experience.

OTOH, I still wonder why the therapist question got so few responses but so many favorites.
posted by sweetkid at 4:44 PM on March 9

This thread is useless without OLAP cubes.
posted by blue_beetle at 5:02 PM on March 9

because if you care about how to stay focussed, you probably have a few tricks on staying focussed

See I look at it totally differently. I bet a reasonable proportion of those that favourited that question did so not because they care about being focussed per se but because they care about the question, i.e. they also would like to know how to do this because they can't. Lots of favourites could even correlate to less answers, because more people don't know the answer but want to. But then I also disagree with this:

There's not one right answer, and every one of us should have a useful tip or three.

I don't see why this particular topic should be something everyone knows how to do. Lots of people never have to stay self-motivated for a start, their lives just aren't set up like that. I also see lots of people around me having trouble focussing and I know that I personally also have a lot of trouble with it. You don't want any of us answering the question, none of us have a good answer. The point of ask.me is for qualified, knowledgeable people to answer each question and it's frustrating when people spout uninformed crap, so I would have thought a lesser number of good quality answers is a good thing

I do remember your question from when it was posted and I also know exactly why I didn't bother reading it. You put a small sentence on the front page which had no question in it and which told the reader nothing. Mystery meat questions like that always get less answers because they, well, kind of suck actually (there have been metatalk posts complaining about this in the past), and I'm personally surprised you got as many good answers as you did. What was there in your statement to entice that well qualified person with an awesome answer to click through and find out what you were wanting? In future if you state your question succinctly and clearly right up front where it can be seen then your answers will come.

Basically, a poorly written question won't be saved by posting at the optimal time of day, and a good question will always get good answers. So look to yourself first.
posted by shelleycat at 5:13 PM on March 9 [2 favorites]

this does get asked a lot, and is a valid, tho oft ridiculed question. My observation, AskMe is USAcentric, specifically East Coast timezone, white collar. I'd wager the best time to ask a question, for the greatest number of sober answers, is between 10 AM and 11:30 on Monday thru Thursday.
10:27 is optimal.
On Tuesdays.
posted by dawson at 5:17 PM on March 9 [1 favorite]

Scattergraphing comments vs favorites intrigues me.
These are unlabelled. The # of comments is the x axis, 0-200; favorites are the y axis, 0-200. 0, 0 at the lower left.

Mefi.

Askme.

MeTa.

Music.

To preserve a bit of detail along the bottom, the color is the remainder of the count divided by 64 - the colors repeat.
posted by Pronoiac at 5:52 PM on March 9

On average you lose a comment or two by posting to askme in the (server's) wee hours, especially midweek, but the variation is comparable to the measurement noise.

The number of comments on a post follows a power law. For Ask posts the exponent doesn't change depending on when you post. For Metafilter posts it might. That would make e.g. a giant trainwreck much more likely during the day.

Pronoiac, neat.
posted by fantabulous timewaster at 7:03 PM on March 9

We're just not that into you, pseudostrabismus.

I have reverse 911 set up. All MeFites that haven't responded to one of my questions get a phone call at 0300 local time asking why the hell not.
posted by lukemeister at 7:36 PM on March 9 [1 favorite]

Average comments per question, by day and hour of asking. Pretty much no change.

Average comments per question, by day and hour of asking, 2008 questions only. More variation, since the sample size is smaller, but still pretty flat.


The code (someone please find a bug in it):

perl -MDate::Manip -lane 'next if (/^\D/);
  my $t=UnixDate("$F[2] $F[3]", "%w%H"); $n{$t} += $F[4]; $p{$t}++; }
  for my $t (sort keys %n) { printf("%.1f,", $n{$t}/$p{$t}); ' postdata_askme.txt

(Note that Date::Manip is obviously slower than substr(). This took 2 minutes to run on my Core 2 Duo T7300.)
posted by Plutor at 5:13 AM on March 10

Can someone please point me to a location of the postdata_askme.txt file? I would like to try this for myself.
posted by grouse at 7:05 AM on March 10

You can grab it from a temporary mirror on my website. Normally it would be hosted on mefi, but we're retooling the perl-on-mefi stuff for security reasons at this point.
posted by cortex at 7:46 AM on March 10

It's available on cortex's mirror of January's infodump that Pronoiac linked to.
posted by Plutor at 7:46 AM on March 10

Damn nosy kids.
posted by Plutor at 7:46 AM on March 10

Maybe you can make a graph of YOU BEING TOO SLOW.
posted by cortex at 7:54 AM on March 10 [1 favorite]

Plutor, I got basically the same result for the mean number of questions: the variation is smaller than the noise. I used Date::Parse instead of Date::Manip, and made some histograms to look at the probability a post gets a certain number of comments, and it really doesn't seem to matter.

If you look at the mean comments-per-post data with some grid lines marking midnight, for all the sites, you might be able to convince yourself there are some features. For Metafilter posts, there's a pretty clear diurnal variation. For music posts, there's sort of a gradual buildup during the week to Friday night. Ask posts have a feature that might be a dip just after midnight on Wednesdays and Thursdays. But you have to squint to see any of this.

The probability that an Ask post gets a certain number of comments follows a power law, and the odds of getting more than a dozen comments doesn't seem to depend on what day (middle) or time (bottom) you post. The same is true for metafilter posts, as far as I can tell.
posted by fantabulous timewaster at 11:26 AM on March 10 [2 favorites]

I tip my hat to you, sir.
posted by Plutor at 11:37 AM on March 10

Fantastic stuff.
posted by cortex at 11:39 AM on March 10

Awesome.

Should we look at medians or standard variations?
posted by Pronoiac at 12:03 PM on March 10

I'm guessing you would get more meaningful results by removing human relations questions. They all get tons of comments and swamp out any otherwise meaningful variations in other types of questions.
posted by vacapinta at 12:23 PM on March 10

There's a statistical test you can do to ask whether two ensembles were drawn from the same probability distribution. That would give guidance on whether differences in other statistics are meaningful, but it would be a lot of work. More interesting than the mean or median is the exponent on the tail of the distribution.

I guess it's not a power law (which would look striaght on a log-log scale, rather than semilog) but some sort of exponential or log-normal or something. Shows what I know about distributions.

vacapinta, interesting suggestion.
posted by fantabulous timewaster at 12:55 PM on March 10