Running the Numbers on Second Life

There’s been discussion lately about Second Life and how its reported numbers seem to be off. Clay Shirky, on Valleywag, has been deconstructing the numbers claiming that the emperor wore no clothes.

The funny thing is that this was coinciding with some research I’ve been doing to better understand whether SecondLife is a flash in the pan of whether it holds real meat. Since October, I’ve been tabulating the numbers listed on their front page, once a week, on Mondays. From there, I ended up with the following spreadsheet:

Date Total Residents Logged in last 60 days US$ spent in last 24 hours Lindex Activity last 24 hours
11-Aug-06 493,563 225,028    
22-Oct-06 1,082,664 446,153 $460,979.00  
24-Oct-06 1,110,224 459,062 $519,914.00  
31-Oct-06 1,203,213 499,223 $531,768.00  
6-Nov-06 1,269,019 515,907 $613,793.00  
14-Nov-06 1,391,715 533,825 $548,912.00 $86,659.00
20-Nov-06 1,517,480 611,793 $656,020.00 $91,508.00
27-Nov-06 1,653,272 667,645 $624,537.00 $99,675.00
4-Dec-06 1,791,247 700,303 $654,750.00 $116,785.00
11-Dec-06 1,932,418 701,287 $626,187.00 $132,005.00
18-Dec-06 2,002,617 720,010 $696,210.00 $120,198.00
25-Dec-06 2,107,321 831,653 $721,341.00 $124,723.00
1-Jan-07 2,287,108 844,317 $803,790.00 $116,425.00

But the data itself wasn’t that interesting when it came to raw form. So I started thinking about some of the things I could do with it. Calculating weekly growth rates was the first thing I looked into but, as more financial data became available, I also started looking at how US dollars to Linden dollars moved along. From there, I ended up with the following spreadsheet:

Date Total Residents Logged in last 60 days US$ spent in last 24
hours
Lindex Activity last 24
hours
Total Residents added
since last check
Increase in 60 days
logins since last check
Increase US$ spent in
last 24 hours
Increase Lindex Activity
last 24 hours
% of residents who logged in the last 60 days $US spend by resident
11-Aug-06                 45.59%  
22-Oct-06 119.36% 98.27%     589,101 221,125 $460,979   41.21% $61.99
24-Oct-06 2.55% 2.89% 12.78%   27,560 12,909 $58,935   41.35% $67.95
31-Oct-06 8.38% 8.75% 2.28%   92,989 40,161 $11,854   41.49% $63.91
6-Nov-06 5.47% 3.34% 15.42%   65,806 16,684 $82,025   40.65% $71.38
14-Nov-06 9.67% 3.47% -10.57%   122,696 17,918 ($64,881)   38.36% $61.70
20-Nov-06 9.04% 14.61% 19.51%   125,765 77,968 $107,108 $4,849 40.32% $64.34
27-Nov-06 8.95% 9.13% -4.80% 8.92% 135,792 55,852 ($31,483) $8,167 40.38% $56.13
4-Dec-06 8.35% 4.89% 4.84% 17.17% 137,975 32,658 $30,213 $17,110 39.10% $56.10
11-Dec-06 7.88% 0.14% -4.36% 13.03% 141,171 984 ($28,563) $15,220 36.29% $53.57
18-Dec-06 3.63% 2.67% 11.18% -8.94% 70,199 18,723 $70,023 ($11,807) 35.95% $58.02
25-Dec-06 5.23% 15.51% 3.61% 3.76% 104,704 111,643 $25,131 $4,525 39.46% $52.04
1-Jan-07 8.53% 1.52% 11.43% -6.65% 179,787 12,664 $82,449 ($8,298) 36.92% $57.12

Some data became clearer as a result of this. Here are a few key findings:

  • On average, the number of logins over a 60 day period seems to be about 35 to 40 percent of the total population reported
  • The people who log in, however, seem to spend a fair amount of money ($50-60 a week) within the Second Life economy. This seems pretty impressive to me. Now, I’m now accounting for the fact that this is not a standard distribution (meaning that some people may not be spending a single dollar in the world) but it seems to point to large amounts of US dollars coming into the Linden economy.
  • The amount of dollars spent seems to vary greatly from week to week with some week seeing a drop compared to the previous week but the averages seem to be going up.
  • Except for a drop around the christmas season (I’m assuming because people had better things to do), it seems the average number of logins no a week by week basis is increasing (based on the 60 day average)

The data seems to support claims of growth relating to Second Life. But how far can it grow? To do that assessment, I decided to make a few assumptions

  Total Residents Last 60 days login
Latest Numbers 2,287,108 596,631
Lowest Growth Rate 2.55% 0.14%
Highest Growth Rate 9.67% 15.51%
Average
Growth Rate
7.06% 6.08%

For this data, I decided to take an approach where I would have a conservative projection (based on the lowest week on week growth number), a liberal one (based on the highest week on week growth number), and a most like estimate (based on an average of all the numbers I had). I then baselined against the most recent reporting period. From there, I now had three scenarios for population growth and 60 day logins. I started estimating it out over the next few months:

  Population Growth Rate Projections   60 day Logins Growth Projection
  Low High Average Low High Average
1/1/2007 2,287,108 2,287,108 2,287,108 596,631 596,631 596,631
1/8/2007 2,345,328 2,345,328 2,448,585 597,470 689,143 632,929
1/15/2007 2,405,030 2,405,030 2,621,463 598,309 796,000 671,436
1/22/2007 2,466,252 2,637,562 2,806,547 599,150 919,426 712,285
1/29/2007 2,529,032 2,892,577 3,004,699 599,992 1,061,990 755,619
2/5/2007 2,593,410 3,172,247 3,216,840 600,835 1,226,660 801,589
2/12/2007 2,659,428 3,478,958 3,443,959 601,679 1,416,863 850,357
2/19/2007 2,727,125 3,815,324 3,687,114 602,524 1,636,558 902,091
2/26/2007 2,796,546 4,184,211 3,947,436 603,371 1,890,319 956,972
3/5/2007 2,867,734 4,588,765 4,226,138 604,219 2,183,427 1,015,193
3/12/2007 2,940,735 5,032,433 4,524,517 605,068 2,521,984 1,076,956
3/19/2007 3,015,593 5,518,997 4,843,963 605,918 2,913,036 1,142,476
3/26/2007 3,092,357 6,052,605 5,185,962 606,769 3,364,725 1,211,982
4/2/2007 3,171,075 6,637,806 5,552,108 607,622 3,886,451 1,285,717
4/9/2007 3,251,797 7,279,586 5,944,105 608,476 4,489,074 1,363,938
4/16/2007 3,334,574 7,983,418 6,363,778 609,331 5,185,139 1,446,917
4/23/2007 3,419,458 8,755,301 6,813,082 610,187 5,989,135 1,534,945
4/30/2007 3,506,503 9,601,813 7,294,107 611,044 6,917,795 1,628,329

So, it looks that, under the most conservative growth rate, we will see 3.5 million users registered and over 600,000 using the service by the end of April 2007. Under a liberal interpretation of the data, those numbers would shift to 9.6 million and just under 7 million. However, in the most likely case, it is probable that there will be 7.2 million users registered with 1.6 million logging in over the previous sixty days. Not too shabby. For the sake of planning, I would advise my readers to go with the most conservative estimate because my data set is still relatively small. Even then, this type of growth mirrors some of the growth patterns we’ve seen in the early days of the commercial web and seem to support the contention that LindenLab is going to be a very strong player in the future.

About the Author

Tristan Louis

Writing and working on the internet since 1993, I've launched six companies, of which two went public and three were sold. This is my personal site and all opinions here are mine.

  • Clay Shirky

    But there is no published metric of “Registered Users” — the Residents figure includes people who signed up but never logged in, as well as alts. “Residents” represents a 50% inflation over actual people, and most of them bail after the first month.

    David Kirkpatrick of Fortune just got some real figures out of Linden, and only 1.5M people have ever logged in (the gap between 2.3M Residents and 1.5M people is from the alts and sign up/no login people), and of that figure, only 250K have logged in more than 30 days later.

    In other words, the highest possible figure for active users is 250K — the actual number is lower, but we don’t know how much lower. It also means that only a minority of the Last 60 Days logins are active users.

    http://many.corante.com/archives/2007/01/04/real_second_life_numbers_thanks_to_david_kirkpatrick.php

  • http://www.tnl.net/blog/ Tristan Louis

    Clay,

    Thanks to your complaints and David Kirkpatrick’s efforts, we have more numbers to ground ourselves in. However, there are two other things I find interesting:
    1. The revenue figures: If you’re correct then my numbers in terms of dollars spent per user are actually on the low side, which would point to an even wealthier crowd which is much more financially active than we’ve come to expect. That, in itself, could warrant some investigation.
    2. The growth rate is the other thing that we should take a look at. Assuming that the numbers of resident they report are inflated, let’s assume that we half the growth rate numbers. So that would be a growth rate of roughly 3 percent a week! Through the magic of compounding, this represents a fairly staggering yearly growth rate.

    So the question is: considering that the “resident” numbers were inflated, should we completely dismiss Second Life or should we look at the other metrics before we actually claim the emperor has no clothes?

  • Clay Shirky

    We should of course look at the other numbers, and there’s no reason to “completely dismiss” Second Life — there are thousands of active users, which means it obviously has some value. Its just a different value, calculated via social proof, than having millions of users.

    To take your questions in reverse order, we don’t know what percentage of users in any given month have accounts that are more than a month old. In December, the highest that number could have been is 250K (which is a measure of anyone who has _ever_ logged in in two months, even if it was last July and August and they never came back.)

    The actual number of current users with a month of experience — call it X — is smaller than 250K, but we don’t know by how much. This unknown number matters because it is the population measure of active users a newbie or business could expect to encounter in SL and establish any sort of relationship with.

    One possibility is that once someone comes back twice in two months, they’re hooked, so growth in the “ever logged in twice” figure tracks X closely. This is Kirkpatrick’s hypothesis.

    Another view is that abandonment continues after the first month, at lower than the 85% one-month rate, but high enough to make a difference. (At a guess, the 3 month activity figure is a third lower than the 1 month figure.)

    So Kirkpatrick got us a snapshot of growth in potential active user base — we do not yet have a second data point to measure change in growth (the 23% figure he quoted was from the month of the Business Week cover) and we do not know how closely active user base tracks potential user base.

    To your second point, about economics, since currency is a big percentage of the transactions in SL, and since abandoning avatars leave their $Ls in-world, some of the economic activity may be a transfer of in-world currency from the absent to the present. (I’ve been talking to someone running a bank in SL, where a chunk of their revenue comes from just such abandonment.)

    If you are looking at economics, drop me a line — [Tristan’s note: Email address has been removed to avoid spamming] — and I’ll introduce you to some of the folks I’m talking to.

  • http://www.tnl.net/blog/ Tristan Louis

    Clay,

    I agree that the percentage of users in a given months is something we don’t know (which is why I started tracking reported data so I could get a more accurate view). However, what we do know is that There is growth in the number of logins. If I read the stat correctly, the logins in the last 60 days representing unique accounts that have logged in during that period. That’s why I tried tracking percentage of users who have logged in the last 60 days. That number seems to be around 40%, which would point to an abandonment rate of 60%. That last figure could be considered high but it appears that it’s not affecting SL than much due to the fact that their new users are coming in at a fast clip.

    So the analogies could be along the lines of what happened to AOL in the beginning or what happens to mobile phone companies. The churn rate is high but there are still people joining in.

    I wish I knew when the data Kirkpatrick had is from and what was the reported number at that time. It’s true that there is a large delta between what Linden reported publicly and what it appears to have reported to Kirkpatrick so Linden Labs needs to come clean on this. The provisioning of historical data would also be an important thing (as far as I know, this post is one of very few to show historical data on Linden Lab at a time when the company could provide that for free).

    As far as economics are concerned I see your point. My view is actually that there is another thing to take account which is the number of times a Linden dollar is recycled. It may be that a single dollar is responsible for 50 transactions a week (and therefore opening up the specter of online gambling and Second Life) so once again, more transparency would be nice. However, to your note about dollars staying in as a result of abandonment, there’s also the question of dollars going out. Based on the Lindex activity, it’s not hard to think that there’s about US$1 million leaving the game on a monthly basis (my own assumption here) and so there could be some interesting effects there. I’ll drop you an email separately on this to get contact info from people looking into this or you can point them to this thread so we can keep it all in the open for everyone else to read :)

  • http://emeraldfalls.blogspot.com/ CC

    Linden Labs numbers are being bloated by the use of resident created bots. I own just one sim and have banned over 150. If this is any indication of the *numbers* then everyone is being fooled.

    Apparently LL doesn’t allow it because they did delete over 100 bots with greek first names in December. Now the bots have profiles and group names and are hard to distinguish from other residents.

    I would rather see on Linden Labs front page the actual number of paying residents or residents with payment info on file. All this hype over their growing numbers is hilarious…but then Linden Lab is laughing all the way to the bank…

  • http://secondthoughts.typepad.com Prokofy Neva

    You need to keep your eye on the number of those who buy premium accounts who own land on the mainland. I think it’s a good number to watch to see if the world is really growing. Add to that the number of new islands bought and number of island owners. The tier-payers and island-bill payers are the people who provide 70 percent of Linden Lab’s revenue, they are like another venture capitalist, taken as a group, spending more than $10 million a year. This figure isn’t over 40,000 people right now, but they account for most of the high end of the economic transactions.

  • http://www.raphkoster.com Raph

    So, am I mistaken in thinking that the $50-60 is just a measure of liquidity? I give $5 to you, you spend it, that person spends it, the next person spends it, and it looks like $20 in transactions. That’s what those figures are measuring. It in no way means that all of that is revenue flowing into Linden or (net) into the hands of users. It’s velocity, not quantity.

  • Morris Vig

    Agreed with Prokofy Neva. The premium accounts are critical.

    In addition, it would be worthwhile to track the number of unverified accounts. An indvidual griefer attack (a malicious user, using an unverified account for the purposes of anonymity) can create a handful of unverified accounts in a single attack.

    An example for the uninformed: A griefer logs on to an unverified account and launches a malicious, self-replicating object at an event. SL residents advise Linden Labs staff, who banish the account. Griefer, having been banned, reenters SL with unverfied account #2. SL residents again advise Linden Labs, who in turn bans the griefer’s new account. And so on. A given griefer attack – like one I personally witnessed this week – can create 5 new unverified accounts before the griefer apparently gets bored and leaves.

    Now, unverified accounts also are used for more productive aims. But I would hazard to guess that many of the accounts where users are NOT logging back in are griefers.

    As a side note on griefers, their attacks are increasing in Second Life. I suspect that control of these attacks will be a major focus of future software upgrades – especially as more and more corporate (read: big dollar spending) customers are coming on line and Linden Labs cannot afford any egg on its face just as it appears to have hit the tipping point.