Log in

No account? Create an account
pinkie pie


Rich Lafferty's Journal

(mendelicious mendelusions)

Previous Entry Share Flag Next Entry
Measuring LiveJournal growth
kernel panic, geeky

In this lj_biz car crash discussion, halfjack wondered what sort of effect the removal of invite codes had on the rate of account creation at LiveJournal. I did a quick gnuplot chart of the result, which was pretty striking.

I decided to do a better chart as a quick project to learn gnuplot a bit better, and ended up charting account creation rate and userbase for what I believe are the five largest LiveJournal-based journal sites: LiveJournal, DeadJournal, GreatestJournal, Blurty, and uJournal.

The most interesting chart is of the number of new accounts created each day:

Graph of new accounts created per day

So invite codes definitely have an effect -- you can see the dramatic changes when LiveJournal introduced and removed invite codes, and smaller but still striking changes on uJournal. (I'm not sure what caused the drop in DeadJournal account creation halfway through 2002; I remember a dataloss issue there around then, but that shouldn't affect new users. Does anyone know when DeadJournal introduced invite codes?)

One thing to keep in mind when reading the above chart is that LiveJournal's decision to abandon invite codes coincided exactly (and intentionally) with the first screenings of the LiveJournal commercials. Some of the peak was obviously the result of moviegoers seeing those commercials, but the level has stayed pretty high since then.

I decided that one huge inline chart was enough, but you can also see the chart of total users per site in the same time period (based on the number of new accounts, which is the only archival data available). You can also see the gnuplot program which created the charts.

All of the data was obtained from the "newbyday" entries of the sites' raw statistics. You may also enjoy reading the equivalent human-readable statistics.

  • 1
Ok, now chart the change of how often people are actually using their journals!! =) I don't think that has changed much, actually, where per-hour is concerned. The chart is pretty rough, right? What I find funny is that my friend gave out his LJ codes only to those that were willing to comment on his entries, and of course, to those that would actually update. Not so oddly enough, his comment stats haven't really gone done. But he's a good writer =P

Unfortunately, historical data of active user count isn't available -- only the current number of active users (for a few values of "active") is available in the stats page.

The chart there isn't rough, though -- it's plotting the exact number of new accounts opened every day since 2001-01-01 on all of those sites (and the linked one, despite those smooth curves, is plotting the number of accounts created since inception for the same period.)

Or I could be misunderstanding what you mean by "rough".

Re: mendel Expand
DeadJournal introduced invite codes in mid-late 2002. I think it was within the first few months of my doing support there, and i started doing support in june 2002 i think.

I could probably a) do research, or b) ask Frank (owner of DJ, not the goat ^.^) the next time I catch him online.

The first mention of invite codes I saw in New_Shit was August 10th. The data loss occured in June.

But around that time, there were lots of other server/hardware related problems that had deadjournal running very sluggishly, which probably contributed to the drop in users.

Re: medlir Expand
Cool! You may want to post or link to this in lj_research too. :)

Tufte would kick your ass for that second chart.

Well, yes, that's why it's not inlined. It was the first chart (save for the one you saw in lj_biz though, and it seemed silly to discard it outright.

I'd be curious to hear what specific bits repulsed you, though :-)

Re: mendel Expand
Wow, cool. News to me! (We don't do pretty stats like this....)

A little Perl and gnuplot can give you some very tidy images for canned periodic reporting. You can even wedge in fairly sophisticated curve fitting and projection without a ton of work. One of my favourite jobs as my user icon probably attests to. :)

Re: mendel Expand
this is fascinating, if not disconcerting. that's an astonishing rate change. I had no idea. how does it translate into message traffic?

I wish I knew! Unfortunately, the historical postsbyday statistics on all of the sites except GreatestJournal dry up in the summer of 2003.

(no subject) (Anonymous) Expand
Re: mendel Expand

that number is an amazing change.

not to say that there isn't the interest... but, can that rate of growth can be maintained? i mean it seems to already be dropping a good deal. before invite codes lj seemed to be approching a non-linear curve. it will be interesting to see what sort of curve develops now.

post by day count is always approximatley the same or grows slowly. that rise of accounts numer is clearly due to people creating multiple accounts.
Current mood:чукча-мыслитель

  • 1