I wanted to see how Metafilter has grown since its launch

January 12, 2001 12:18 PM

I wanted to see how Metafilter has grown since its launch, so I crawled the archives for thread/comment totals for every day since July 1999. Then, I wrote a script to plot all the data in line graphs and tables. Click on the links to drill down.

It's interesting to see how quickly the commenting community have grown. At this rate, Matt will have to build thread-within-thread support to handle the sheer number of comments.
posted by waxpancake to MetaFilter-related at 12:18 PM (16 comments total)

That is absolutely rockin'! :)
posted by sonofsamiam at 12:33 PM on January 12, 2001

It's interesting how the stats went down during December. Could it be all the October/November extra traffic was election (and post-election) related? Nawww.... :)
posted by pnevares at 2:50 PM on January 12, 2001

Woah election peak! Well, there was at least 3-4 days of extended downtime in December, my pageview stats are low for that month too.

waxpancake, can you plot threads on a separate axis? The statistician in me knows that the recent rise in posts/day is obvious, but not when it is set against an axis that goes to 5800. How about a 0-50 axis?
posted by mathowie at 3:13 PM on January 12, 2001

Holy cats. What happened in December of 1999 that caused postings to nearly double the next month? Growth was sort of flat until then, then BOOM!
posted by Skot at 3:55 PM on January 12, 2001

Yeah, no problem Matt. It's easier to do two separate axes than to have a single mixed graph. I'll make the changes later today.
posted by waxpancake at 4:13 PM on January 12, 2001

I took wax's numbers and plotted comments per thread[*] . To me it looks like there have been three eras in the history of MF:

1. Matt solo: Matt posts a link, nobody responds (through Dec. 99)
2. Transition: As more people discover MF, comments/thread skyrocket (Jan-May 2000)
3. Saturation: Number of comments plateaus (June 2000 on).

I think there's a simple overload factor at work here: as threads get long, people stop being able to follow them (especially if this happens really fast, as in yesterday's "Kurt Cobain" thread. I got home from work, found 80 comments, fuhgeddabouddit) so they post new (often redundant) links instead. So you get N discussions, not very different, spaced out over about N days, instead of one long, focused discussion.

*(actually, [comments + threads]/comments, because I wanted the main link itself to count as a comment, so a link with no comments = 1)
posted by rodii at 4:28 PM on January 12, 2001

Skot, in mid January, the site got a ProjectCool award, and the pageviews went from 100 per day to 6 or 7 thousand, and although it tapered down to 1-2 thousand after that, there was enough people to get discussions going. It was a good critical mass sort of growth.
posted by mathowie at 4:39 PM on January 12, 2001

You can now switch between one or two charts on the main page, Matt. If anyone has any other ideas for data-crunching, let me know.
posted by waxpancake at 5:05 PM on January 12, 2001

I think December numbers dropped because of the holidays, too. Lots of people weren't able to pay as much attention as they normally do because of being away from computers and what not.
posted by cCranium at 5:12 PM on January 12, 2001

Wax: I'd like to see: all these things compared against number of members, number of active posters, comments per thread per unit of time, and the Dow Jones and NASDAQ indexes. And sunspots. :)
posted by rodii at 7:05 PM on January 12, 2001

Actually, thanks to rcade, I'm working on a query to determine a "discussion creation ratio" where I figure out the ratio of number of comments posted to threads you have created over the number of threads you have created.

Now, if you weren't good at posting provacative things, I'd expect your number to be 0-5, and I think some members might top out at 20-30.

It's a good metric, except for the ability of someone to post something very shocking, that creates lots of comments, and they might not necessarily be good.
posted by mathowie at 10:42 PM on January 12, 2001

Actually, re: the discussion explosion, I looked carefully at the January data and even before I read the thread one in particular stood out. As Matt points out, it was the first to go past 10 comments. I think I have now proven the A list conspiracy.
posted by norm at 4:24 PM on January 15, 2001

That or the fact that porn really is the Internet's killer app. :-)
posted by cCranium at 4:38 PM on January 15, 2001

Boy, it's really nice to be able to wake up an older thread like this instad of starting a new one. Are we sure that MeFi shouldn't work more like MeTa?

Anyway, my question is: Brother Wax, would it be hard to modify that script of yours to do a day-by-day count? It really feels like the number of new threads each day is getting to be rather overwhelming, to the point where I'm starting to overlook them entirely. I'm too lazy--or maybe I would just feel like too much of a geek--to go back and do some counting, but that's what computers do best, right?

The Dec. figures say about 13 new threads a day. today it's 18 so far, 16 yesterday. . . why does it seem so high?
posted by rodii at 5:05 PM on January 18, 2001

If you click on a month, you'll get a day-by-day breakdown of threads. For example, this page breaks down all of December. (Am I misunderstanding you?)
posted by waxpancake at 11:28 AM on January 19, 2001

No, you're not, I'm just a dork. Thanks.
posted by rodii at 1:02 PM on January 19, 2001