Thread/Comment and User Growth statistics, July 2001

August 2, 2001 8:40 AM

Thread/Comment and User Growth statistics are up for July. This is the first month that threads and comments actually dropped slightly, largely because new user signups are still disabled. Also, we broke 10,000 user IDs yesterday. Congratulations, Ricci.
posted by waxpancake to MetaFilter-related at 8:40 AM (26 comments total)

User growth has a nice little curve to it. It seems like growth has decelerated, not levelled. Is this a function of disabling new user sign ups? Are there no more people out there who want to chit chat?
posted by rschram at 10:03 AM on August 2, 2001

Threads: Comments Ratio, or mean comments per thread

July 1999: 1.90625
August 1999: 0.239130434782609
September 1999: 0.282051282051282
October 1999: 0.211764705882353
November 1999: 0.484536082474227
December 1999: 0.345679012345679
January 2000: 2.03821656050955
February 2000: 3.66094420600858
March 2000: 8.45514950166113
April 2000: 5.73556231003039
May 2000: 9.20403022670025
June 2000: 8.01690821256039
July 2000: 9.63855421686747
August 2000: 7.91282051282051
September 2000: 9.83838383838384
October 2000: 10.4247967479675
November 2000: 10.448087431694
December 2000: 11.3507462686567
January 2001: 14.9455782312925
February 2001: 17.1810154525386
March 2001: 16.7416520210896
April 2001: 16.9385826771654
May 2001: 19.6299342105263
June 2001: 20.1217391304348
July 2001: 21.1369230769231

posted by rschram at 10:49 AM on August 2, 2001

i wonder what the ratio would be like if you were to exclude a few exceptions, such as the Kaycee threads and the one thread (i can't remember which) that has something like over a thousand comments on it.
posted by moz at 10:57 AM on August 2, 2001

Moz, I don't really know. I think it doesn't really has as big an impact. The infamous 1142 is an outlier, but it was so long ago, it doesn't change much our analysis of trends over the last 12 months.

I am trying to capture the proverbial "signal to noise." I am interested in seeing whether the ratios of thread to comment is accelerating or decelerating. By looking over the ratios, late 2000 and early 2001 have very rapid rise in average comments. More recently the rise has been less rapid.

To me, this means that, while comments are far above threads, and continue to grow with added users, the number of posts per month is reaching a kind of saturation point where the number of posts is too many for people to process.

If you think about it, this is plausible because the addition of new posts impacts people's browsing of previous ones. Hence, a small increase in posts over time will have a big impact on average comments, even if comments continue to rise more or less exponentially.

What this indicates for me is that, notwithstanding growth in users, MeFi's "daily pace" posts is also growing.

It would be more useful to look at the change over time of the difference in rates of change of threads and posts. I think that's the same as what I've done with the ratios, but I'm not so good at this math stuff. It would also be more direct to see a curve for the ratio of posts/day to comments/day, since that goes to "daily pace."

There's a paper in this, I can smell it.
posted by rschram at 11:32 AM on August 2, 2001

i have noticed for some time now that certain posts get ignored after a point. if you post something early in the morning, and you come back in the early evenings, the threads are dead. done and over with. i don't know if that's been the case at MeFi for a long time now (i joined up in late April).

maybe this is obvious, but it seems like what attracts people to comment on threads is activity. if there is no activity, few bother. i don't know if the links in the threads themselves are ignored, or if they're followed but not commented on, but it seems that the fear of being alone in saying something about a thread is important to a lot of mefi users.

it isn't so much a problem that a thread is not commented on -- the notion of "interesting" links often getting few comments has been kicked around. what i'm curious about is whether people decide to click through to a link based on the number of comments it has. thoughts?
posted by moz at 12:06 PM on August 2, 2001

maybe this is obvious, but it seems like what attracts people to comment on threads is activity. if there is no activity, few bother. i don't know if the links in the threads themselves are ignored, or if they're followed but not commented on, but it seems that the fear of being alone in saying something about a thread is important to a lot of mefi users.

Another possibility is that a significant percentage of users are using the "recent comments" view and don't see the less-trafficked posts.
posted by snarkout at 12:18 PM on August 2, 2001

Me, I switch between the 'date' and 'recent comments' views. I do think saturation is one aspect here: at a certain point, I don't feel I can add anything signalacious to a thread. I'd rather not be snarky or flip, given this problem, but I think that's often peoples' reaction to the thread overload problem. I know I preview and cancel nowadays almost daily, where I never would have bothered to before.
posted by dhartung at 1:03 PM on August 2, 2001

I ran some new calculations. The results are mixed, and I would appreciate others' analyses.

This shows the change from the previous month in average comments per thread.

Launch 0-July 1999: 1.90625
July 1999-August 1999: -1.66711956521739
August 1999-September 1999: 0.0429208472686733
September 1999-October 1999: -0.0702865761689291
October 1999-November 1999: 0.272771376591874
November 1999-December 1999: -0.138857070128548
December 1999-January 2000: 1.69253754816388
January 2000-February 2000: 1.62272764549903
February 2000-March 2000: 4.79420529565255
March 2000-April 2000: -2.71958719163073
April 2000-May 2000: 3.46846791666986
May 2000-June 2000: -1.18712201413987
June 2000-July 2000: 1.62164600430708
July 2000-August 2000: -1.72573370404696
August 2000-September 2000: 1.92556332556333
September 2000-October 2000: 0.58641290958364
October 2000-November 2000: 0.0232906837265094
November 2000-December 2000: 0.902658836962727
December 2000-January 2001: 3.5948319626358
January 2001-February 2001: 2.23543722124611
February 2001-March 2001: -0.439363431448999
April 2001-May 2001: 2.69135153336096
May 2001-June 2001: 0.491804919908468
June 2001-July 2001: 1.01518394648829

This is meant to show more clearly the change over time in the average number of comments per thread. Waxy's graph gave me an impression that comments rose much faster than threads, as if each thread were more substantial and longer as well as there being more to choose from.

As you can see, MeFi has had changes in both directions of avg comments/thread. Last month, comments dropped a bit because of disabled new user creation.

This suggests that, although we are hovering at around 20 comments per thread, the consistent rise in the rate of thread-posting each month does not always return in equally strong growth in comments.

However, this does not appear to be a recent development. As such, I don't see persuasive evidence of a trend toward posts outstripping comments.

It is worth noting that this change in average comments does not correlate with the exponential growth in new users between the latter part of 2000 and the early part of this year. So basically user # 9000-10000 are looky-loos, not Chatty Cathies. But then there could be a flaw in how I conceived the whole thing...
posted by rschram at 1:27 PM on August 2, 2001

it's always seemed to me that the more vocal members of the mefi community are those who have fairly low numbers. why?

it might be because they've been here longer, and are therefore more comfortable. or it might be that they're from a generation of mefi where, due to the low number of users, you had to be speak up if you wanted there to be any sort of activity on a thread. or it could just be that a lot of the users who come to metafilter now are less concerned with joining and participating in a community than they are with checking out the site and seeing what's going on -- specifically, many may have simply joined just to chime in or stay abreast of the kaycee thing, but their attention may have waned.

of course, all of those justifications sound about as right as old baseball adages like "put the ball in play" or "pitchers do worse with men on base" or some such.
posted by moz at 1:58 PM on August 2, 2001

[T]he more vocal members of the mefi community are those who have fairly low numbers. To test that out, we'd have to look at user data--correlate user birthday to post/comment ratio. That doesn't quite get to the issue of silence on the part of new users, but it's somewhat relevant.

As an experiment, I drew 10 random numbers between 1-10,000. I recorded their posting stats from the user page. Only 3 of 10 had ever posted anything, or commented on anything. Of that, 2 users posted 1 link each, and one comment. These users ran the gamut of ages.

There's a lot of dead wood out there.
posted by rschram at 3:26 PM on August 2, 2001

almost makes you wonder if we shouldn't recycle user #s.
posted by moz at 3:29 PM on August 2, 2001

I am trying to capture the proverbial "signal to noise."

If you're going to write that paper, you're going to have to do a lot of work to convince me that posts/thread is in any way correlated to signal/noise.
posted by anapestic at 3:40 PM on August 2, 2001

If you're going to write that paper, you're going to have to do a lot of work to convince me that posts/thread is in any way correlated to signal/noise.

The idea of a signal is obviously subjective. You or I might not think that posts about who won the election are "signals," but so far they have consistently had lengthy if not insightful discussions. Meanwhile other perhaps more thought-provoking or "quality" posts get no attention.

I take the view that comments are votes, and that long threads are thus the strongest signals. From the point of view of the people who do comment in those threads, and indeed, prolong them, the rest is "noise". It's not in any way a value judgement on the posts. It's a way to assess the impact of MeFi's growth on many fronts to the length of discussions. Any further judgements would requrie rereading those threads! No way!

Signal/noise may not be the best terms for the question I want to ask. I want to know what the relationship is between growth and people's ability to find what they on MeFi and contribute to it.

posted by rschram at 4:09 PM on August 2, 2001

Do I win a prize? 10,000 by..... August How long before 20K?
posted by brent at 5:19 AM on August 3, 2001

Questions:

*How do you determine what member number you are?

*Can you render statistics on number of posts by member number? Because I believe that you'd find a very low number of Chattie Cathy's, perhaps between 200-300, if that, which leads to the question...

*If you only lurk, and you can lurk without becoming a member, why would you become a member?

*Can you determine activity per member number in such a way that recycling member numbers is a viable thing to do? Perhaps saying, if you haven't posted in 60 days, or you don't respond to a query email in 60 days, your membership number goes back into the hat...?

*If you can recycle numbers, won't that make the cachet of having a low number meaningless? Or would everyone move up? As in, I'm #10000, two people get booted and their numbers recycled, and my number subsequently becomes #9998....?
posted by UncleFes at 8:22 AM on August 3, 2001

*How do you determine what member number you are? You click here. It's in the URL.

*Can you render statistics on number of posts by member number? ...

Each user page has a count of comments and posts. Now how to get that information conveniently is another issue about which I know nothing.

*If you only lurk, and you can lurk without becoming a member, why would you become a member?

Good question. Maybe a whole ton of people signed up for the 5k. Maybe it's not new users lurking as it is old users leaving Metafilter. There is no way for a user to delete a userid.
posted by rschram at 9:34 AM on August 3, 2001

It's in the URL.

Ahhhh....dork city, baby :)

Now how to get that information conveniently is another issue about which I know nothing.

conveniently is the kicker. Anyone *could* go one by one and log the link/post numbers, but urg what a job.

But I would bet that, if the engine tracks the posts for reviewing, somewhere there's a database holding all that info. Where's there's a populated database, there exists the ability to generate a report.

Which leads me to a more culturally-based question: if you are a Mefi member, do you have an obligation to post at least occasionally? What about me - lots of posts, but no links? Should members be required to participate, or renounce their membership?


posted by UncleFes at 9:49 AM on August 3, 2001

There was a discussion in Metatalk awhile ago about the comment/post ratio as a moral barometer. The higher your score, the better a person you were because you "gave back" to threads and discussions.

Of course its wrong on two counts: 1, If you have no links and 1000 comments of "Bush sucks" and "Florida stole the election" then are you really better than anyone. 2, My C/P ratio is 3.3. Surprisingly low, and yet, I'm widely loved and admired. ;)
posted by rschram at 10:12 AM on August 3, 2001

*If you only lurk, and you can lurk without becoming a member, why would you become a member?

I lurked for a loooooong time before I started posting (as you'll see if you view my userpage), and I lurked for a while before I registered. the reason I finally did register wasn't because I was ready to start participating, but because I wanted to customize the front page, plain and simple.

so that's at least one answer to your question. :)
posted by rabi at 10:16 AM on August 3, 2001

Is there a way to measure/monitor how many distinct viewers there are for a given thread (as opposed to contributors)?
posted by ParisParamus at 11:05 AM on August 3, 2001

Is there a way to measure/monitor how many distinct viewers there are for a given thread (as opposed to contributors)?
posted by ParisParamus at 11:05 AM on August 3, 2001

I'm with ParisParamus ... a lot of the threads that I enjoy the most are the ones where my viewpoint has already been stated ... so I don't comment. That doesn't think that I think that thread is "noise" ... so maybe the page views would be a better way of judging signal vs. noise.

posted by Johannahh at 1:05 PM on August 3, 2001

There's no way to know without trying but I think comments and views correspond. Johannahh makes the point herself: [A] lot of the threads that I enjoy the most are the ones where my viewpoint has already been stated ... so I don't comment. A thread with many comments receives proportionately many page views.
posted by rschram at 1:10 PM on August 3, 2001

Rscham, my question was aimed at determining how many people read Mefi; and individual threads. The longer a thread, the more likely someone's viewpoint has been expressed. There's actually a negative relationship between thread length and the percentage of viewers. In other words, a thread with 35 contributions may have been read by 400 people; a thread with 40 posts may have been read by 600.
posted by ParisParamus at 8:01 PM on August 3, 2001

I think that the lurkers are necessary. I myself lurked for at least 6 monsths before registering. The reason I lurked was because MeFi has become my news source. It's a sure bet that anything of importance will show up and I will get to read an extended point-counterpoint. The more trivial front-page posts take the edge off of constant political ramblings and make it entertaining.
I registered for two reasons-
1.) to make a comment on a thread on something I had a personal stake in a while ago
and
2.) to have something to do when I got home from the bars and had nothing better to do.

Lurkers exist because some people are simply not into discussions where every possible point has been made. MeFi is much more than just a discussion site. It's also a great utility site and a helluva lot more fun than CNN.
posted by ttrendel at 11:30 PM on August 3, 2001

Someone has to say it about our new user #10000: Ricci, don't lose that number, it's the only one you own.
posted by kindall at 9:18 AM on August 23, 2001