How many active users are there?

In the spirit of this Hotline question, I'm reposting my analysis of Vasudha's numbers. Feel free to (intelligently) criticize my analysis or post your own estimates.


I'm going to make a modest leap here and try to calculate the number of distinct active basic accounts.

35% of all premium accounts who logged in in March made a LindeX purchase. For the record, this is consistent with the purchase activity we saw on GOM (recall this was prior to the alt-happy days of the free basic accounts). 8% of basic accounts that logged in in March made a purchase. Of course, some of these basics are alts. The unknown quantity is how many basics are either alts of another basic, or of a premium account.

Vasudha's post states that 70% of the currency was purchased by premiums, and 30% by basics. If we assume that *making a purchase* is independent of account status, with only the *amount* varying, then we come to an estimate of:
# distinct basic accounts = 47,000 basics logged in * 8% making purchases / 35% of logged in users purchasing = ~10,800 distinct, active basic accounts

This would put the total active SL population at around 25,000. Given a daily peak usage of around 5,000 users, we're looking at a peak concurrency of around 20%. This is on the high side with respect to other games, but is consistent with industry estimates.

If you take the average ppl online at any moment = 4500 (a conservative average estimate), and the average (!!) time spent online per day in SL is 5 hours (I think the average time online might be way way less), you get about 21000 users who log in during one day. I think the average time spent online per day is 4 hours or less, that means 27000 users who log in during a day.
There are many who don't log in every day, just every few days, so the active ppl count might be much more higher... I'd estimate 30-50000 active users, who log in at least once a week.

I did a concurrence test by capture-recapture on my friends online list - and got someone who (bravely or insanely has never deleted any calling cards) to do the same.

We both got about 20% concurrence from the numbers that we crunched.

Since Active Player Count is more important than total player count,
Linden Labs should stop harping on a 150,000+ user base. Know darn well
that many of those folks haven't logged in for months/years.

The User Base should be scaled as such (See Below). This would be a fair accounting
of the players subscribed in Second Life. Because LL has (NO) policy for the
deletion of inactive accounts. Yet they count those accounts as valid population
numbers. Kind of like if the US Census started counting people born in 1776
as current members of the US 300/million population tally. If the Census didn't
remove dead people, the US population would be over a billion people. Nevermind
that 70% of them are dead.

Little issues like this makes you wonder about the accounting Linden Labs
acquires, crunches, and uses in its economic policies in SL.
Personally, SL's Active population is more like 10% of the total population number.


0-7/day Active User Count
8-28/day Active User Count
29-90/day Active User Count
91-180/day Active User Count
181-365/day Active User Count
366+day Active User Count

I do have to agree with RBD on this one. He does have me wondering now about cancelled accounts. Are they counted as part of the 150,000+ or not? If they are.... that is kinda, sorta, maybe false advertising in a sense. Nothing they can't get away with legally, but, it's the principle of the matter that makes it wrong in my book. If an account has been cancelled it shouldn't be a part of the population count any longer.

If you actually cancel an account (ask LL to get rid of it), I believe that it no longer counts as part of the resident count (194.787 when I just checked), even though it still exists somewhere in the database. (For instance, if you have a calling card for a former resident, you can still see their profile information. They don't show up in Find People, though.) However, many people just abandon their accounts rather than cancelling them, and those still count as part of the number.

We do have the amount of unique logins for each month from the data LL released. If somebody has kept track of the total amount of accounts you can deduct the active accounts and the trend.

Mar '06 = 60.899
Feb. '06 = 57.812
Jan. '06 = 56.462
Dec. '05 = 47.898
Nov. '05 = 44.836
Oct. '05 = 43.683
Sept. '05 = 34.609
Aug. '05 = 26.473
Jul. '05 = 25.639
June '05 = 21.554
May '05 = 18.593
April '05 = 18.324
Mar '05 = 16.184

In the previous on-line business I had some involvement with, the stats worked out at around 5% to 10% of registered users were active, that is prepared to spend money.

With Second Life it is a bit more complicated, that is due to the alt accounts and the basic stipend.

I presume it may have three main classes of users

1) Premium Account holders

2) Alt Account holders, some with and some without Premium Accounts

3) Basic Account holders

Within (3) there are possibly abandoned accounts

However it is again possible quite a few Basic Account holders are economically active due to the stipend or land rental rather than ownership

Here's a start at a second method of estimating the number of active accounts...

We assume there are N total active accounts. This is the quantity we wish to find. The information we have is the average login time per user, A, the maximum number of concurrent users online, C, and the total number logging in in a given month, K.

Assume a distribution of logins, f(t), with t in [0, 1] (i.e. the fraction of the day that has passed when the user logs in). If logins were independent of time, this would be a constant function. Given that most SL users are located on the west coast, with the second largest population coming from the UK, this function should actually be roughly bimodal with a periodicity of 24h.

Assume also that a person stays logged in for a length of time L each day. L is in the range [0, 24h], with an average of A. I would guess that L would be Beta distributed. Given the mean and variance of the login times one could calculate the distribution parameters.

Given the login time and duration distributions, one can estimate the probability that a certain user is logged in at a certain time. That would be given by:
P(logged in at time t) = P(logged in prior to time t && login duration greater than (t - t_login)) + P(logged in after time t && login duration greater than (24h - t)). Note that in order to filter out alts and count only truly "active" accounts, one should place a lower limit on the login time.

With an assumed distribution of login times, it should be straightforward to calculate the distribution of login probabilities as a function of time. By dividing the peak number of concurrent users by probability for the peak time, one can obtain an estimate of the number of active accounts.

The algebra is left as an exercise to the reader.

I think your calculations are unnecessarily complex, need way too much measured information, and they won't give much more precise results. In fact, due to much more variables needed, a variation in one variable can cause some shift in the result.

I still prefer my way, or the 'concurrency percentage' way (the latter is the easiest, no doubt).

BTW, the concurrency percentage and the 'average time spent in SL per day per user' are interdependent, both can be calculated from the other.

20% concurrency means that an SL person spends an average of 4.8 hours in SL per day.

I don't think this is correct. The concurrency figures quoted for online games refer to "peak concurrency" -- the peak fraction of users online at once. So, picture a situation where nobody is online all day. Then 20% of all users log in at once for 1 minute then log out again. This situation will produce a peak concurrency of 20%, but the average online time would be nowhere near 4.8 hours.

Correction: Linden Labs has no PUBLISHED policy for the deletion of inactive accounts.

They do delete inactive accounts, unless the handful of returning players I've talked with were lying and really cancelled their basic accounts prior to going inactive.

Whether or not the culling process is manual, automatic, who knows. But where there are rules there is policy, written or not.

Oh, I thought you meant average concurrency. This means the average time online per user is less than 4.8 hours, meaning more active users.

Not that average concurrency is THAT hard to measure using the same methods: for a few days, you could check the number of friends online (provided you have many many friends on your list), every 10 or 15 minutes. This could be done using the SL website 'friends online' list and a simple script. Then you could get the 'average concurrency' rate, if several ppl. did this measurement.

But in the end, I think you'd arrive at the same figure: about at least 25 - 30 000 users are logging on and are active during the course of any one day.

Yes that is consistent with what I see but.....

Are those 25/30k users each unique logins or are they mainly alt accounts sitting in camping chairs.

I suspect we may see an answer to that one with the removal of dwell

Just a personal view

I have a friend who owns 3 premium accounts and one basic account. She paid yearly for the premium accounts (3 X $72) because they offered 3 main benefits:

1. 500L$/week = 26000L$/year X 3 = 78000 L$/year ... - ... ($72x3x275L$/$) = 18600L$ profit

2. She speculates with the 3 X 512 sqm land allocation turning it over 3X each month at an average profit of 1000$ each transaction. 3000L$ X 3 accounts X 12months = 108000 L$

3. She uses the 3 premium accounts to buy group land therefore gaining a 10% benefit on land held by the group.

The value of turning over the 512 sqm of no-tier land and the benefit of the stipend amounts to a healthy 126600 L$ or $422 at today's exchange rates.

I wonder how many others have premium alts who buy nothing, yet cash in L$ every month. I think you greatly overestimate the number of true active users.

I don't think cancelled accounts would be counted --- that's pretty much illogical even to a 3rd grader. But, the numbers might be skewed a bit by people who leave and are just too lazy to cancel. I know that if I leave I might not even bother to go and do the cancel procedure.

L$ is at 300 L$/$ now, but that still means L$ 13000 of profit per year, true.
'Only' risk is the market instability, which could easily mean that she'll be at L$ 10000 or more loss, if the market crashes.



If she's such a good speculator, I'd say that she should get 64 000 sqm of land tier and speculate with that. She'd 'easily' prop up her profits to L$ 4.5 million, or 15 000 USD/year.
But wait, maybe even 15 000 USD/year isn't such a great pay, though.



How many hours of work is needed to achieve this $422 profit? What's the hourly wage? What is the amortization/operational cost of the broadband and the PC used?
What happens if the L$ falls to 360 L$ / US$?
What happens if Linden Labs closes?
Did you account for conversion and cash out fees? (About 3%?)

That 25k-30k figure isn't logins per day. That number is the estimate of the total number of actual, honest-to-god people using Second Life. Anything above that number is comprised of alts, free basics that have abandoned SL, and premiums that don't log in any more.

My point was not to argue about my friend's productivity vs. entertainment value, nor about the minimal time it takes to buy a 512 sqm lot and set a price on it. Nor about using average exchange rate over past year of 275. Nor did I want to get into the broadband being "sunk cost". My point was simply that there are many others out there with alt accounts that contribute little to SL economy except to drain real $.


I don't need a lesson on financial accounting nor in economics from amateurs.

Maybe their point was that this scheme is amateuristic and silly enough for it not to be used by many people. Obviously the stipends are not exactly the big benefit in the scheme. The land is and thus any sane person would simply use a single premium account and tier up. Instead of listening to our amateuristic views on economy and accounting I suggest you go advise your friend a bit on how she can vastly improve her efficiency. Given it's trivial for amateurs like us to find improvements your results can't be less than spectacular.

I think that I would take advice from Mr Q about Second Life (and with respect I do have an understanding of financial accounting/planning and markets)

Which after all from a strict economic view is what Second Life is all about

Huh? Better stick to your day job, I think.

Look, for a single premium account tiered up to any level greater than 512 sq m, the cost of adding and holding each additional 512 is greater than $0, right? Yes, the marginal cost goes down somewhat as tier level increases, but it's always positive and significant, going no lower than US$195/mo for each 65536 (+512).

However, for a group of multiple premium alts, the marginal cost of adding and holding an additional 512 (563 w/ the 10% group bonus) is -18,600 L$ per year! In other words, the more alts you add, the more free tier AND each one has built-in base profit of 18,600 L$ per year.

LL has a policy of no more than 5 accounts per household for a good reason. Otherwise, we'd all have 500, 5000 or whatever we could actually physically manage! Each one contributing a free 563 sq m of tier to the group and each with a negative marginal cost (a guaranteed profit).

You're totally right but you fail to see that most of what you say is besides the point within the example. The whole stipend scheme depends on the L$ rate being less than 348L$/1US$. This risk you take at the beginning of a year and the return comes each month. If the L$ rate gets worse your margin shrinks and you have no way to step out of it as you invested upfront.

On the other hand if you invest 72US$ + tier you've the ability to alter your strategy month by month depending on how well sales go. Once you hit 25US$ of tier you beat out any scheme that uses alts and stipends as long as you have a 1000L$ mark up per 512sq m 3 times a month. When you extend it to 40US$ tier your average mark up or amounts of turnovers can drop quite a bit before your margin is worse than what you would get from the multi-account scheme.

So sane business advice is to to drop to 1 premium account and gradually expand the selling business. That is off course within the boundaries of the example.

My dayjob requires far more complex business advice than that and I feel quite confident I'll do just fine.

If you have the legal max of 5 accounts per real life address that are all basic......

250 Lindens bonus for signing up and 50 sl per week, over one year that works out to
$L 2,850 which is around $10 US give or take

The next 4 accounts cost a one off fee of $10 US each and generate the same number of Lindens per year

So at year end total profit would be plus minus $10

Year 2 onwards would be pure profit on all accounts

Now lets assume the user has fast broadband (UK now can get 8 megs) and a fast computer.

I have been told you can run multipal versions of Second Life (source Michael Linden on telephone help line) so therefore all your little individual avators could in theory hang around on camping chairs having purchased (as an investment) someting to stop themn getting logged out.

Lets say an average rate of $L2 per 10 minutes times 5 thats $L 10 per tem minutes or $L 60 per hour, or $L 1,440 per day or $L 525,600 per annum.

Add back the 2850 per account ($L 14,250) that equals $L 539,850

Or at an average of $L 310 = $ US that would be $1,741.......

I wonder how many powerful Company and University computers are sitting around running multipal copies of Second Life