When should I ask mefi? January 21, 2008 1:24 PM   Subscribe

What is the best time of day (or, even more specifically, day of week and time) to ask metafilter?

Are there particular times of day for posting when one is most likely to get more answers? Or more best answers? Do these numbers vary based on category/topic?

Has anybody got any data about this?

Thanks!
posted by rbs to MetaFilter-Related at 1:24 PM (53 comments total) 4 users marked this as a favorite

Previously, also, etc, and so on.

I'd like to point out that the "etc" link was posted by you.
posted by cortex (staff) at 1:28 PM on January 21, 2008 [17 favorites]


Yeah, I don't think the answers have changed much in 9 months since you last asked it.
posted by mathowie (staff) at 1:33 PM on January 21, 2008


Why is this thread still open?
posted by HotPatatta at 1:34 PM on January 21, 2008


When Mercury is in retrograde and it's snowing in Montana.
posted by jonmc at 1:35 PM on January 21, 2008


#1 can close if he wants, but I was hoping for some hard data.
posted by rbs at 1:35 PM on January 21, 2008


The best time of day to ask is whenever iconomy is around. If the hive mind has a queen bee, I think she's it.
posted by cashman at 1:37 PM on January 21, 2008


#1 can close if he wants, but I was hoping for some hard data.

I think that hard data can be a little difficult because it doesn't really matter if your question gets 102 answers if none of them help you. And some people like to Best Answer everything and some do not. There are numbers we could crunch but they're going to wind up at the same place which is that most people seem to hit MeFi during the roughly 9-5 US workday range, not so much on weekends and lighter on weekend nights. It's not Ebay, I really don't think you can do ROI numbers for it.
posted by jessamyn (staff) at 1:43 PM on January 21, 2008


The best time of day to ask is whenever iconomy is around.

I'm pretty sure that for a certain fellow whose name rhymes with Bolden, the worst time of day to ask is whenever Miko is around.
posted by scody at 1:50 PM on January 21, 2008 [1 favorite]


Now, if anybody really wants to do some crunching, here's the askme post stats including datestamp and number of comments. Doesn't include Best Answer information, but a raw analysis that assumes number-of-answers as an explicitly positive metric might be interesting.

rbs, I think a quantitative analysis would be fun, and there's a few people who might do it, but I don't think there's a really high quality, verifiable answer to be found. The general speculation folks have put forth here and previously is probably as close as we can get to a rule of thumb.

Of course, if anyone would like to volunteer to read through 1.2 million answers and score each of 'em on a 1-5 scale, I'll happily post-process that.

Anyone?
posted by cortex (staff) at 2:01 PM on January 21, 2008


Anyone?

Cooter?
posted by Brandon Blatcher at 2:03 PM on January 21, 2008


Cooter?

Wait until I go for my PhD, at which point I'll do the work and then it will be DOCTOR Cooter to you reprobates.
posted by jessamyn (staff) at 2:05 PM on January 21, 2008


I think the grocery at the bottom of the hill has cans of that stuff in the cooler. It doesn't taste very good.
posted by Wolfdog at 2:25 PM on January 21, 2008


(I need a phrase that's like "oh snap!" but less irritating. Dang. Well, insert one here.)
posted by aramaic at 2:27 PM on January 21, 2008


Looks like you may have picked the wrong time to ask this.
posted by weapons-grade pandemonium at 2:32 PM on January 21, 2008 [3 favorites]


You're no Congressman Cooter!
posted by dersins at 2:37 PM on January 21, 2008


Just wait 'till SCDB is awake - You'll get your answer.
posted by davey_darling at 2:38 PM on January 21, 2008


Wait until I go for my PhD

Cortex, why can't you be more like Jessamyn?
posted by Brandon Blatcher at 2:39 PM on January 21, 2008


I prefer to fancy myself Honorary Doctorate material. One of these days, somebody'll notice my inimitable contributions to humanity and BAM! Dr. Butts Lol.
posted by cortex (staff) at 2:45 PM on January 21, 2008


Jessamyn, let me know if you're going for a Ph.D., rating all the answers on AskMe. Let's just say I know a little bit about that. Or something related.
posted by needled at 2:49 PM on January 21, 2008


Oh, don't be coy, needled. Dish.
posted by cortex (staff) at 2:56 PM on January 21, 2008


I figured you're one of the first people I'll ask, needled :)
posted by jessamyn (staff) at 2:57 PM on January 21, 2008


Wow ... I wonder what else hides in that stuff.metafilter.com subdomain that us peons can't see?

I know, I know ... you'd show me, but you'd have to ban me.
posted by scblackman at 3:03 PM on January 21, 2008


Wait until I go for my PhD

I'd study it.
posted by Brandon Blatcher at 3:08 PM on January 21, 2008


One of the days, I'll put together a nice little index, scblackman.
posted by cortex (staff) at 3:09 PM on January 21, 2008


Just wait 'till SCDB is awake - You'll get your answer.

Strangely enough, an anagram for "Steven C. Den Beste" is "Cents Been Vested" and lord knows he loves to put his two in
posted by jtron at 3:11 PM on January 21, 2008


cortex, let's just say it sounded simple on paper, until we got the ratings in. And no, this wasn't Ask MetaFilter.
posted by needled at 3:14 PM on January 21, 2008


Strangely enough, an anagram for "Steven C. Den Beste" is "Cents Been Vested" and lord knows he loves to put his two in

You ended a sentence with a preposition! Where's languagehat when you *really* need him?
posted by wafaa at 3:17 PM on January 21, 2008


I'm not sure that The Hat is batting for the team you think he is, wafaa.
posted by cortex (staff) at 3:19 PM on January 21, 2008 [1 favorite]


Carter, D. S., & Janes, J. (2000), "Unobtrusive Data Analysis of Digital Reference Questions and Service at the Internet Public Library: An Exploratory Study," Library Trends, 49(2): 251-265.
posted by RichardP at 3:26 PM on January 21, 2008


Jessamyn: If you decide to do that as an actual dissertation (one day), it'd be interesting to compare your ratings (and reliability of them against some rater you train) to what the community might do if we were all able to score stuff (many hands make light work). Even more interesting would be to know about the topical distribution of answers by contributor (e.g., the people who really like to answer the health related questions) and how that is related to the characteristics of the answerer (maybe you could conduct interviews). Y'all have a huge dataset here...
posted by rbs at 3:28 PM on January 21, 2008


The best way to get the most answers is to make your question about cat behaviour. The second best way is to ask how to improve a hopeless relationship.
posted by happyturtle at 3:32 PM on January 21, 2008 [1 favorite]


Some stuff.
posted by timeistight at 3:35 PM on January 21, 2008


Hmmm... Looks like google only has the Word Clouds and the Markov Filter.
posted by timeistight at 3:38 PM on January 21, 2008


"The best way to get the most answers is to make your question about cat behaviour. The second best way is to ask how to improve a hopeless relationship."

How can I be sure my cat really loves me? [more inside]
posted by mr_crash_davis at 3:48 PM on January 21, 2008 [4 favorites]


Data is fun. The first anonymous question was nine months before the 2nd.
posted by smackfu at 4:17 PM on January 21, 2008


jessamyn, you'll always be Dr. Cooter to me.
posted by scrump at 4:31 PM on January 21, 2008


There might be some revisionist (gasp!) action going on there, smackfu; we've relatively recently gained the ability to easily make a question anonymous after the fact, and get a request now and then from someone experiencing major regret months/years after the fact.

So really, you probably want to look for the first group of anony questions to find where they really started—or look for timestamps that jump forward and back again against a numeric sort of the thread ids, since true anony submissions reserve a thread id and sit in a queue until they're posted.
posted by cortex (staff) at 4:40 PM on January 21, 2008


Well, cortex's numbers were easy enough to manipulate, and I was bored enough, that I ran some basic stats on them. I don't think there's all that much to learn from the comment counts just from what I did, but there were some interesting things about favorites. I computed mean by hour, and by day. 0 is Monday and times are PST (I assume, I didn't change it from the raw data). I haven't thought at all about how viewers from different time zones will affect analysis of the hourly data.
 hour   avg. comments   stdev    avg. favorites  stdev
 0:00           13.87   11.78              2.00   5.88
 1:00           13.34   12.29              2.14   5.95
 2:00           13.64   13.26              1.69   5.18
 3:00           13.86   12.95              2.02   5.29
 4:00           14.11   13.96              2.20   11.47
 5:00           14.37   14.35              1.93   4.89
 6:00           14.44   14.52              2.04   5.48
 7:00           14.46   14.06              2.03   6.15
 8:00           14.18   13.44              2.06   6.79
 9:00           14.19   13.15              2.07   7.41
10:00           14.54   14.05              2.01   5.91
11:00           14.26   13.82              1.93   5.92
12:00           14.13   13.88              1.83   5.48
13:00           14.01   13.21              2.02   6.60
14:00           14.14   13.43              2.07   6.63
15:00           14.04   13.34              1.91   6.03
16:00           13.98   12.51              1.99   5.24
17:00           14.55   13.76              2.41   15.41
18:00           14.79   18.41              2.17   6.61
19:00           14.59   13.42              2.21   7.26
20:00           14.44   12.96              2.13   6.70
21:00           15.07   15.85              2.36   7.38
22:00           14.85   14.67              1.98   4.68
23:00           14.74   14.80              1.94   4.62

day   avg. comments   stdev    avg. favorites  stdev
  0           14.23   15.45              2.01   6.27
  1           14.54   14.10              2.05   6.67
  2           14.55   13.71              2.02   6.79
  3           14.39   13.74              1.99   5.93
  4           14.49   13.54              1.94   5.75
  5           13.93   13.07              2.23   11.67
  6           13.76   13.07              2.21   6.78
For comments on questions, it seems that you might be slightly best off posting around 9pm PST, but the stats are not that revealing. Perhaps a day x hour chart would be more interesting (I am not bored enough to do it though). The higher standard deviation at 6-7pm indicates that there are more posts leading to a lot of comments made then, though this doesn't translate into the mean in any real way. So it might mean that if a post has the potential to have lots of comments, and is posted around this time, it is more likely to reach its potential.

The favorites are more interesting. As far as I can tell, the places where there is a high standard deviation for favorite counts represent times where there is more likely to be a sort of "favorite storm", assuming I didn't make a mistake in the calculations. So this tends to happen on posts made on Saturdays, and at 5-6pm and 4-5am. It's hard for me to tell how much the variance has been affected by just a few outliers, but eyeballing the data, it looked like there was a real pattern for at least the Saturday data vs. other days. The 4-5am data is highly influenced by a 343 favorite outlier (in particular, this one, which has even gotten another favorite since cortex pulled that data). The 5-6pm data is also influenced by an 839 favorite outlier (this one), but in general seems to have more outliers (and more posts) than the 4-5am set. Presumably the huge outliers also consist mostly of favorites made later than the post time.

The high variances at these times might be an indicator of how much people really get into a post at different times of day. This seems very likely for the Saturday data, and would make a certain amount of sense for the late afternoon, but the hourly bins might be too small to draw this conclusion. The 4-5am high standard deviation simply looks like an accident to me. It is interesting that the high standard deviation in favorites for Saturday does not correspond to any higher average number of comments, and even in terms of avg. number of favorites it isn't that different from sunday. Perhaps one interpretation is that it's not that people are favoriting a huge amount more on Saturdays, but that there's more of a community effect on Saturdays where people flock to the same posts.
posted by advil at 4:45 PM on January 21, 2008 [3 favorites]


Nice, advil. The relatively uniform distribution here doesn't give any hint of it, and so it might be a goose chase, but I'm wondering what an examination month-by-month would turn up—do trends change with the season? With a heavily US population, does the change in northern hemisphere daylight patterns get reflected in site activity? Etc.
posted by cortex (staff) at 4:58 PM on January 21, 2008


What is the best time of day (or, even more specifically, day of week and time) to ask what the best time of day (or, even more specifically, day of week and time) to ask metafilter is?
posted by signal at 5:01 PM on January 21, 2008


I wrote a crappy perl script. It calcs the ratio of well-answered questions (> 5 answers) to all questions for each hour, then scales the result so it goes from 0 to 20, then prints that # of stars. (I took enough college statistic to now this is "not good", but I still wouldn't ask a question at 2 AM.)

00:00 = **************
01:00 = **
02:00 =
03:00 = *********
04:00 = **************
05:00 = ***********
06:00 = ************
07:00 = *************
08:00 = **************
09:00 = *************
10:00 = ************
11:00 = ***********
12:00 = *************
13:00 = **********
14:00 = ***************
15:00 = *********
16:00 = ***************
17:00 = *************
18:00 = ***************
19:00 = ***************
20:00 = **************
21:00 = ******************
22:00 = ********************
23:00 = ***************

posted by smackfu at 5:14 PM on January 21, 2008 [1 favorite]


Is that Eastern or Pacific time?
posted by clevershark at 6:34 PM on January 21, 2008


Good question. It's whatever is in the file. Probably PST?
posted by smackfu at 6:36 PM on January 21, 2008


So clearly the answer is: ask your question at 10 PM on a Monday in September of 2005.
posted by languagehat at 6:42 PM on January 21, 2008 [2 favorites]


The best time to ask will be when the last Ask Mefi question is asked, because that will trigger the Metafilter Singularity.
posted by lukemeister at 7:32 PM on January 21, 2008


I'd like to tag on a question of my own: what's the best time to ask MetaTalk "what's the best time to Ask MetaFilter?"?
posted by AmbroseChapel at 7:57 PM on January 21, 2008


Three hours ago.

Well, nice to see this question keeps going and going.

(I'm the "also" link above.)
posted by intermod at 8:30 PM on January 21, 2008


Many of the "best answered" questions are answered in the first comment correctly. If you could separate chatfilter from actual questions, you could make a more meaningful analysis.
posted by roofus at 2:36 AM on January 22, 2008


So I'm thinking maybe an alternate view of the data that includes additional "best answer count" and "category" fields, some time soon.
posted by cortex (staff) at 6:30 AM on January 22, 2008


This is awesome.
posted by rbs at 7:14 AM on January 22, 2008 [1 favorite]


I smiled when I saw this thread went way beyond the third comment, and actually provided answers. High five, rbs: you have numbers!
posted by not_on_display at 4:41 PM on January 22, 2008


Sorry for suggesting this thread be closed, rbs. Now I'm just pissed that it took 5 MeTa threads to finally get the question answered.
posted by HotPatatta at 10:58 PM on January 22, 2008


Wednesday, 3-4 PM.
posted by madman at 8:00 AM on January 23, 2008


« Older Frenetic in concert in Cambridge, MA   |   Yay, first place! (but not for long) Newer »

You are not logged in, either login or create an account to post comments