SecondLife Statistics (Updated)

Raph Koster has some good thoughts on the numbers game of accounts in SecondLife. I was writing a comment that ended up being pretty long, so I'm rephrasing it here.

From my perspective, there are different ways to look at the numbers. The number of accounts in SecondLife and the number of actual humans involved are obviously different - there are alternate accounts which are being used for various reasons, from anonymity to running businesses to everyone's favorite topic, griefing¹. The statistical analysis crunches different in more than a few ways, depending on what they are being used for²:

(1) People in world want to know how many people are actually participating, for business or sheer geekiness, which validates what you are doing.

(2) Linden Lab has to always try to scale (and it is having problems) for the maximum amount of users, so it is conservative for them to work with the Big Numbers that we all distrust. The reality, for them, is that the architecture needs to be able to support more people (not accounts), and a reasonable trend they (IMHO) should be using would be the ratio of residents online at the same time to the number of accounts. I suspect that this ratio has increased a bit, but will remain fairly proportionate per capita. Publishing that ratio would be worthwhile.

(3) First world corporations who only use SecondLife are more interested in potential reach (maximum number of people) versus the maximum number of accounts. However, the maximum number of accounts is not the maximum number of people, therefore statistics for reach are (as everyone knows) inflated in that regard. Weekly unique statistics would be better for potential reach as well. However, first world corporations haven't demonstrated that they understand that the real numbers that they should be interested are probably the traffic numbers on their land. To create a useful statistic from SLurl links, one would have to compare web page traffic to traffic on the area where the SLurl links to, and even then it has to be judged based on what the expected effect is.

(4) Mainstream media likes big numbers because people are desensitized to small numbers. 1 million residents is something that they can write about; 15,000 people online at the same time pales in comparison. This actually works against (2) because it creates more free accounts which may not be 'quality' accounts, and can scare the hell out of someone at LindenLab when they crunch the numbers without knowing how many alternate accounts are there. This inflates (3) which in turn gets (1) more active...

The crux of the problem is that there is no number for 'alternate accounts', and in fact creates a case for 'free accounts' to be terminated - but it also decreases the chances of people entering the world, trying it out and staying. At this time, since there are free accounts, it would seem that the belief in a conversion from free to 'payment information on file' accounts is high. Whether Linden Lab has internal numbers to support this would determine if they are acting on 'warm fuzzy' or real numbers.

With Zee Linden on board, I suspect that if they don't have numbers, they are in the process of getting them. And if you consider that it is technically possible to ban someone on account information, it's apparent that there is a way to tell whether there are duplicates of the same information within Linden Lab that may not have been data mined yet or which are simply not being made public.

Speculation on all of this can lead down many paths. No one will know until the numbers are found and released.

Update: Micropersuasion points to some statistics which are worth noting - and also points out economy statistics are more useful, which I missed above.

¹ Since no one knows how many alternate accounts there are, there is no statistical basis on saying that free accounts have resulted in more griefing. Therefore, anyone who says that free accounts create more griefing are, in fact, stating an opinion that they cannot substantiate reasonably.
² Just like any other statistic.