June 22, 2006

5:31 PM

Here are two charts about metafilter. Posts and user numbers and comments and user numbers. Each dot represents a post (or comment), with the post number and number of the user who posted (or commented) as the xy data. I don't remember what prompted me to make these (some metatalk thread, I'm sure), but I found them again on my hard drive and thought someone else might find them interesting (I brought them back up to date, too). There are half a dozen interesting little milestones that show up in the graphs, more than I could see probably.
posted by Lirp to MetaFilter-related at 5:31 PM (64 comments total)

awesome!

So I can see from the comments graph especially that the more vertical the rise as you go from left to right, the more heavy signups and commenting was happening. So for the Nov. 2004 time I turned signups on for the first time in a year, there's a huge vertical spike as people flooded in and commented and posted threads.
posted by mathowie at 5:40 PM on June 22, 2006

Also if you overlay them, you can see the one week waiting period for the November 2004-present people. The posts fall just a little bit behind the comments wave.
posted by mathowie at 5:41 PM on June 22, 2006

This is too cool.
posted by rhapsodie at 5:42 PM on June 22, 2006 [1 favorite]

Broken.
posted by arcticwoman at 5:54 PM on June 22, 2006

not broken.
posted by cellphone at 6:08 PM on June 22, 2006

I always find graphs of stuff like this quite interesting and both links work for me.
posted by Captain_Science at 6:09 PM on June 22, 2006

Oh neat, you can easily pick out 9-11 from the line of comments going up.
posted by cellphone at 6:09 PM on June 22, 2006

Any clue what caused that spike mid-2002?
posted by Serial Killer Slumber Party at 6:14 PM on June 22, 2006

Oh, also, this is cool.
posted by Serial Killer Slumber Party at 6:14 PM on June 22, 2006

Nice nerd porn.
posted by graventy at 6:26 PM on June 22, 2006

you know what's amazing about this? the horizontal banding in the graph. you can see precisely what user numbers were most prevalent in either graph at any given time.

it looks like the 14kers just can't shut up, apparently. damn them!
posted by shmegegge at 6:30 PM on June 22, 2006

So what happened to kill off most of the 5000-7000ers?
posted by Orange Goblin at 6:55 PM on June 22, 2006

Yeah, (what OG mentioned) is the most interesting thing in the graph, I think.
posted by Ethereal Bligh at 7:03 PM on June 22, 2006

Spectacular.

One thing I need help with though. On the comment graph, thereare data points attributed to user numbers early on in MeFi's life, when it appears that there were not that many users at that point. Specifically, what is going on with that vertical line around the comment # 1000. The way I interpret that is that there were always about 15,000 users and all of a sudden they started to comment and then they stopped all of a sudden...but that doesn't make sense when you consider the general trend that the graph illustrates.

Is this clear? Am I missing something obvious? If so, please don't tease me...I tend to cry a lot these days.
posted by Pacheco at 7:08 PM on June 22, 2006

Vertical lines like that are cult posts from days before threads were closed after 30 days. As new users arrived, they'd learn of the old thread and post to them.
posted by Ethereal Bligh at 7:11 PM on June 22, 2006

Any clue what caused that spike mid-2002?

Me.
posted by yhbc at 7:11 PM on June 22, 2006

The 14kers, and anybody signing up after a long closed gate period, is more consistently posting over the long term. Look at how piddly the staying power is of the folk signing up recently.

It screams out to me "Shut the gates periodically".
posted by Meatbomb at 7:12 PM on June 22, 2006

(So, in other words, although there is a corrleation along both axis to time, it's not a perfect correlation.)
posted by Ethereal Bligh at 7:12 PM on June 22, 2006

Any clue what caused that spike mid-2002?
It was me too--it was the reopening of signups in the summer. : >
posted by amberglow at 7:15 PM on June 22, 2006

It's interesting that one of the densest areas on the posts graph (aside from early on) is my little cadre of April Fools. Not that I've contributed much to that.
posted by Ethereal Bligh at 7:16 PM on June 22, 2006

got it, thanks
posted by Pacheco at 7:17 PM on June 22, 2006

So what happened to kill off most of the 5000-7000ers?

Those were my practice accounts.
posted by cortex at 7:18 PM on June 22, 2006

Also, Lirp, I love you.
posted by cortex at 7:18 PM on June 22, 2006

Hey, there I am! Heh.
posted by sugarfish at 7:33 PM on June 22, 2006

Wow, that is totally sweet. With the 9/11 thread and the early spike in comments... People must have gone back after regstering and posted in really old threads...
posted by delmoi at 7:49 PM on June 22, 2006

Very cool! For those who don't know, there's also Waxy's stats, but they haven't been updated in a while.
posted by stavrosthewonderchicken at 7:50 PM on June 22, 2006

So what happened to kill off most of the 5000-7000ers?

It looks to me like a little under 6000-10,000, but it is an amazing hole, especially on the comment graph.
posted by Chuckles at 7:51 PM on June 22, 2006

Also, I'm guessing that the vertical stripe on the left here must be the legendary 1142, in which pretty much everybody commented. The one further along would be the equally legendary 9622. Neat!
posted by stavrosthewonderchicken at 7:52 PM on June 22, 2006

Actually, I can spot both 1142 and 9622 based on their position within the post numbers. It looks like the stripe on the farthest left is either a graph artifact or possibly number 19.
posted by yhbc at 7:57 PM on June 22, 2006

Ack, nevermind - the line I was looking at designates the year 2000.
posted by yhbc at 7:58 PM on June 22, 2006

Good eye stav. That's cool.
posted by eyeballkid at 7:59 PM on June 22, 2006

Any clue what caused that spike mid-2002?

The reopening of new account signups NOT after long closed period but after a long period where only special people invited by Matt could join. I believe this the era before this is the period that baked the quality into this community. Casual users dropped away (that's how I read the 5-7Kers, and many many 1-10Kers), engaged users stayed on and gained a kind of establishment. The slow feeding in of really cool people only served to supercharge this golden age. Finally the floodgates opened, but people tended to fall into one of two camps: 1) like-minded souls who belonged here in the first place 2) newbs who were willing to respect (or unable to stop fearing) very well-established norms. There were some crops of losers in there from time to time as well but we ate them alive, as we do now.
posted by scarabic at 8:12 PM on June 22, 2006

special people invited by Matt

Or, lurkers who liked the place so much they figured out how to sneak in.

Hi, iconomy!
posted by yhbc at 8:19 PM on June 22, 2006

Those people were also invited by Matt, but not explicitly :)

Ooh, this is cool. Too bad the font tag is now disabled. You know, the font tag would really destroy this place.
posted by scarabic at 8:20 PM on June 22, 2006

You can also see that user posting activity fades more quickly than commenting activity.
posted by gsteff at 8:49 PM on June 22, 2006

Wow, this was more popular than I expected.

One of my favorite bits is the band of totally unused accounts right above user ten thousand. I'm guessing there were some blogs with "metafilter hits 10k users" posts, so a bunch of people found the site, signed up, and promptly forgot about it, maybe?
posted by Lirp at 9:17 PM on June 22, 2006

The spike in 2002 was when I would open and close signups randomly, like two days every few months.

The big dip in activity in certain segments dates back to when the 5k contest shared the same server and I let Stewart use the mefi database as the signup.

Some italian dork entered a few crappy 5kb pages and wrote bots to create hundreds of accounts, along with a few thousand real people that also entered. Those people rarely came over to MeFi to participate. In the early days, MeFi skewed so far into the web development community that having the 5k contest use our user db was convenient and a lot of members here entered.
posted by mathowie at 9:18 PM on June 22, 2006

Also, it'd be cool if you ran the numbers on Ask Mefi, which doesn't really get used by a ton of user id's below, say, 14k or so. I bet that would produce some weird use graphs.
posted by mathowie at 9:20 PM on June 22, 2006

I was thinking about doing askmefi (and metatalk), too, but you could probably do it faster than I could, Matt. I had to gather all the information by hand (ok, by perl script), and I'd have to do it again for the other sections of the site, while you've got it all in a database already. I'll send you an email with the ploticus file I worked up, and you can see what you think.
posted by Lirp at 9:33 PM on June 22, 2006

That's cool. I can see the blip when I was able to sign up in April '04, or at least I think I can. The comment activity of the 14-21k seem to be the most dense, although there is a pretty intense band centered aroun 13k as well.


Nifty Lirp, this should go on the sidebar.
posted by edgeways at 9:53 PM on June 22, 2006

Yeah, lemme send you a simple csv file Lirp.
posted by mathowie at 11:14 PM on June 22, 2006

Jeez, people have been asking for some sort of database dump for years. To get it, all I had to do was spider the whole site first?
posted by Plutor at 5:40 AM on June 23, 2006

Excellent job, Lirp. Any chance of an annotated graph with little 'milestones' highlighted? I am math-impaired but can read some basic English. (Or perhaps we can put it on Flickr and have people highlight their own finds?)

I'd love to see the graph(s) for AskMe, btw.
posted by goodnewsfortheinsane at 6:11 AM on June 23, 2006

Yeah, lemme send you a simple csv file Lirp.

Uh, I'd like to get in line, too.
posted by cortex at 6:48 AM on June 23, 2006

it looks like the 14kers just can't shut up, apparently. damn them!

*makes pointless comment just to support the team*
posted by languagehat at 7:04 AM on June 23, 2006

Happy to see plenty of evidence to be proud of the Nov. 04 $5 n00bs. It was worth the wait.
posted by Miko at 7:16 AM on June 23, 2006

The most interesting thing to me is that, on both graphs, the more vertical the curve on the left (the more quickly people were joining the site), the more comments and posts these people tended to make.

Users who joined where the boundary curve has a more moderate slope, when signups happen more casually, tend to post and comment less.

I suppose this is obvious, that people who wanted to join the site badly tend to use it more than those with a casual interest, but it's interesting to see it verified in graphical form.
posted by driveler at 7:34 AM on June 23, 2006

Any way of posting the raw data sets, Lirp? I've always wondered what the half-file of a typical member is. My gut says about 18 months or so (and those graphs suggest it). It would be cool to do some activity stats on those datasets.
posted by bonehead at 7:56 AM on June 23, 2006

It would be cool to do some activity stats on those datasets.

Yes, it would.
posted by cortex at 8:17 AM on June 23, 2006

So interesting. Thanks Lirp. I signed up randomly in Sept. 2000 when I spotted mefi in my referrer logs. Didn't own a computer till 2005, forgot I was a member (but lurked occasionally) for the intervening years before remembering I had a login and miraculously remembering what it was last year some time.

The patterns in these graphs are cool to look at, is what I'm saying.
posted by Joseph Gurl at 8:26 AM on June 23, 2006

one of the densest areas on the posts graph < is ethereal bligh's> little cadre of April Fools. Not that I've contributed much to that.

If I were adding notation, I'd call that area the EB Band.
posted by sleslie at 8:42 AM on June 23, 2006

It's like 15,000 MeFi users suddenly all screamed in terror, and then we silenced...
posted by blue_beetle at 9:26 AM on June 23, 2006

So freaking awesome. Thanks, Lirp, I love this.
posted by jonson at 9:52 AM on June 23, 2006

Will someone please specify which dot or series of dots represents the ceiling cat thread?
posted by |n$eCur3 at 10:32 AM on June 23, 2006

Amazing Lirp. Thank you. Coolest thing I've seen in awhile.

Are there any web gurus out there who can work this up to have the dots be clickable? Background image for the graph layout, clear and blue one pixel images for the dots? Or one pixel dots via CSS maybe?
posted by jwells at 11:13 AM on June 23, 2006

Ah, 2002.

Bliss was it in that dawn to be alive,
But to be young was very heaven!

I got in just before the gates slammed shut. I'd been lurking on MeFi ever since the kuro5hin.org/Met4filter.com April Fool's Joke of 2001, but had avoided getting a membership until it looked like Matt was about to shut them off for good.
posted by Pseudoephedrine at 3:09 PM on June 23, 2006

If you don't mind my asking: why? A number of other people have mentioned holding off on joining (though in their cases it was usually until it was too late), and I'm just curious why people would be reluctant to join a community they found so interesting. I mean, you coulda been a 14K-er!
posted by languagehat at 4:42 PM on June 23, 2006

If you don't mind my asking: why? ... I mean, you coulda been a 14K-er!

Question and answer, all rolled in to one.
posted by cortex at 4:44 PM on June 23, 2006

Now, that's uncalled for. I'm gonna sic the sarge on you.
posted by languagehat at 5:05 PM on June 23, 2006

I'm guessing the reason for the 14Kers continuing to post so much is that most of us had to beg in or *cough* break in. You don't do that unless you really give a shit.

It would be interesting to see what would happen if signups were ostensibly closed again, but a backdoor was left open for those who could find it. The old one wasn't very hard when you think about it - the 'Sorry, signups are closed' page was called newuser1.cf, and changing that to newuser2.cf got you to the unblocked signup page.

I'd be curious as to how reimplementing something on that level would improve the post/comment quality around here.
posted by Ryvar at 7:19 PM on June 23, 2006

Yeah, lemme send you a simple csv file Lirp.

Please do. It shouldn't take long to do the same thing for askmefi and metatalk if I can get the data.

perhaps we can put it on Flickr and have people highlight their own finds?

Go for it. I don't have a flickr account myself, but consider these images in the public domain and go to town.

Any way of posting the raw data sets, Lirp? I've always wondered what the half-file of a typical member is. My gut says about 18 months or so (and those graphs suggest it). It would be cool to do some activity stats on those datasets.

The data I collected is not very, uh, robust. I was trying to get date and time information, but didn't manage to get most of it, so there's a lot of incorrect and missing date info. All I really have are user numbers and post/comment numbers. Still, send me an email if you want and we can figure something out.
posted by Lirp at 8:26 PM on June 23, 2006

Flickr'd: comments; posts.
posted by goodnewsfortheinsane at 10:47 AM on June 25, 2006

languagehat> It's one of those things where when the possibility that you might not be able to do it in future looms, even if you weren't thinking of doing it, you still want to keep your options open.

Also, more personally, I was more heavily involved in several other sites, kuro5hin.org amongst them, that were going down the dumps and I wanted a new community.
posted by Pseudoephedrine at 6:19 PM on June 26, 2006

Thanks!
posted by languagehat at 5:43 AM on June 27, 2006