userbase has been growing at a near-exponential rate April 9, 2001 3:23 AM   Subscribe

So, it seems like Metafilter's getting pretty crazy lately? This might be the reason: the userbase has been growing at a near-exponential rate. (I wrote a little script to plot the number of users over time to a graph. Hope you like it.)
posted by waxpancake to MetaFilter-Related at 3:23 AM (20 comments total)

That curve looks more sigmoidal than exponential to me. Slow growth, explosive growth, slow growth. It could be leveling off.

In any case, what your measuring is the number of registered id's, is it not? You're not accounting for attrition or for multiple id's from the same user. It would be interesting to know what percentage of the registered users are regular readers/posters and whether that percentage changes over time. There's probably a limit as to how many people can effectively participate in MeFi at a given time, but there may not be a limit as to the number of registered users (or at least that limit may be so high as to be effectively infinite).

Perhaps a user who doesn't log on for a period of time should be removed (given sufficient warning, etc.). Then you might have a more accurate count. And if the user numbers were automatically re-indexed, we could all move up the ladder. It might cause some cut-throat tactics among the MeFi ambitious, but it's a dog-eat-dog world out there.

And, yes, I'm joking. About the automatic re-indexing, at least.
posted by anapestic at 6:57 AM on April 9, 2001

Slow growth, explosive growth, slow growth.

If you're talking about the level off at the end there, I'm pretty sure it's artificial. As of this moment we have 5212 members, so I think waxpancake is counting end-of-month stats for the month. So April can't be decided until it's over. :)
posted by pnevares at 7:50 AM on April 9, 2001

It looks like you're right, pnevares: I should have read the axes more carefully. Blows my analysis right out of the water.

I wonder at what point it will start to level off. It seems like a fertile field of research for some budding academic. And there's probably no shortage of data out there to analyze.
posted by anapestic at 8:17 AM on April 9, 2001

That curve looks more sigmoidal than exponential to me.

Actually, I think that it's a quartic curve. The number for April 2001 only contains 8 or 9 days of data. I fit a curve to the data (number of days vs number of registered logins; starting with Jan 2000=0 days since prior to Jan 2000, there are testing numbers and inexact dating; 1 data point per month) and the quartic curve fit best (judging by the residuals; least squares regression). The equation that results is:

(Y=# of users; X=# of days)
Y = 237.6 + 5.54*X + 0.0136*X^2 + -1.10*X^3 + 2.56*X^4
(see the graph here)

This equation predicts the following milestones (ceteris paribus).

ID#; Date
6000; April 27, 2001
7000; May 17, 2001

Of course, this all assumes that there will be no changes to the pattern of people joining the community when Matt makes those upcoming changes.

Does anyone know if the quartic pattern is a typical one in community growth? Also - I think that quartics figure quite prominently in chaos theory, but it's not even close to my area of expertise.
posted by iceberg273 at 8:39 AM on April 9, 2001

Math geeks.
posted by waxpancake at 11:43 AM on April 9, 2001

Is there any way you can also gather and plot the number of active members during a month? To me, that would be people who posted a post or a comment at least once during the month. Although I realize that many people read without posting and to a degree they are "active" too.
posted by gluechunk at 12:15 PM on April 9, 2001

I can't retrieve/manage that information easily. Every comment from every thread would need to be indexed.
posted by waxpancake at 12:53 PM on April 9, 2001

To me, that would be people who posted a post or a comment at least once during the month.

I can run a query like this with about five minutes of work. I'll try it tonight when I get home.
posted by mathowie (staff) at 1:16 PM on April 9, 2001

Oh - also, I'm constantly asked how many non-member visits the site gets. I've long been considering grabbing data on that.

I figured doing a simple query on the main index page for a user's cookie, then writing the visit to a database, could be a good way to figure out how many member/non-member visits there are a day, and even how many multiple member visits there are, though that last data point would probably scare me (and others) more than anything.
posted by mathowie (staff) at 1:21 PM on April 9, 2001

I can mention it now, since it's gone to press, but after this comes out on Friday, I suspect any and all equations will break for the growth we're about to experience. :)
posted by mathowie (staff) at 2:03 PM on April 9, 2001

Yeah, yeah, the cover's nice. Personally, I'm looking forward to the centerfold. (It's all about where you hold the monitor.)
posted by bradlands at 2:13 PM on April 9, 2001

well, originally, they weren't going to use a 21" montior, but then they sort of had to...

*budadump bump* (rimshot)
posted by mathowie (staff) at 2:49 PM on April 9, 2001

Ahhh... dick jokes. Let's hope that the huge influx of Brill's readers comes here first!
posted by anildash at 3:33 PM on April 9, 2001

Ok, so I checked the number of members that have visited within the past 30 days, and I get 2175. In the last week, 1352 have checked out the site.

Those are higher than I would have thought.

I haven't nailed down the queries yet, but I'll try and tease out "# of members that have posted a comment or link in the past 30 days" soon, though of course those should be less than the number that visited. It'd be nice to know the % of members that haven't posted in the past month, but who have visited. I wonder if it's over or under 50%.
posted by mathowie (staff) at 10:06 PM on April 9, 2001

posted by waxpancake at 5:58 PM on April 10, 2001

registered users v. lurkers would be interesting to know, too.
posted by crunchland at 8:48 AM on April 11, 2001

...the userbase has been growing at a near-exponential rate.

It certainly looks like it is hitting the steep part of the curve. I have noticed gains everyday this week.

posted by quirked at 10:07 AM on April 11, 2001

how about people who are registering more or less just to vote for 5k entries? Are those added to the permanent user-base? I'd imagine this will effect things as well...?

posted by darkpony at 9:00 PM on April 11, 2001

I just joined Metafilter on the 9th (of April), after months of semi-lurking (meaning I didn't visit everyday, but fairly regularly). I haven't posted much (I think this is number 2) cuz I don't have a lot to say - yet.

I might have an ID from almost a year ago, but I couldn't tell you what it was. I bet there are lots of IDs like that.

Just a perspective from a less regular poster.
posted by melissa at 10:09 AM on April 16, 2001

I updated the user graph... The curve is getting really steep, and we're only halfway through the month.

Note that the difference between the user count on the front page and my user count is due to more than just latency between updates. The number on the front page is just the highest user number... I'm not counting deleted or invalid accounts, like this one. There were several of these in the beginning of Mefi.
posted by waxpancake at 3:33 PM on April 17, 2001

« Older Can't actually POST after PREVIEW   |   TimeZone Bug Newer »

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