This inspired me to try to understand the business model behind a multiplayer game with non-traditional revenue (i.e. not subscriptions). Not just understand, as in … offer game free to play then sell extras … but to really grok (gawd I hate that term) the finer detail.
I probably could have picked a company and game that was a little easier to understand, but I figured I’d take a stab at understanding the business that is Second Life.
There’s always the question of whether SL is a MMOG. My response: who cares. People use it … it’s not a necessity of life … and it has non-traditional revenue streams: in short, if it’s not a MMOG, it’s a close enough analogue that it doesn’t matter.
Second Life provides a wide range of useful data: land, registered users, 60d unique active, etc. You can find the data here.
IMPORATANT NOTE: I must disclose that I’ve never played SL. This hasn’t stopped anyone else from commenting on SL’s business model, so I figure I might as well proceed. I’m NOT associated with Second Life nor Linden Labs. The figures I present are guesswork, based on assumptions and some limited data released by Linden Labs. There may well be errors in the calculations. This is presented solely as an exercise in understanding the business model of a MMO. Draw your own conclusions.
I’m also new to SL’s business model. If anyone can see any problems with what I’ve done here, please comment. More generally, I’d be appreciative of comments (or pointers to articles) on non-traditional business models for online games. Fun facts and figures about retention, conversion, concurrency etc, are also greatly appreciated.
First thing I tried to do was understand SL’s active users numbers. Their main page (as you can see above) gives out the much-maligned ‘registered users’ stat. Here, it shows some 4.8 million.
While others prefer “weekly unique users”, I figure 60 day unique is fine. I’ve had subscriptions that I haven’t logged into for two months. Ideally, of course, you’d measure users by their intention to “play” in some future time period. Short of hiring a fortune-teller, this isn’t practical. Having been “in-game” once in the previous two months is as good an indicator of your intention to return as any other (especially for a casual MMO).
I could only find 60d unique numbers back as far as 2006/05/05. So I took a shot at modeling user growth. Each month, new users will arrive, old ones will leave. An article on GigaGamez suggested that SL retains 12 – 15% of users beyond 1 month. For sake of argument, I assumed another 5% would leave before the end of 3 months, and the rest would be gone before the end of a year.
On these assumptions, I came up with the following chart.
The pink data to the right is actual 60d unique user counts reported on the SL stats pages.
Interesting to note that if growth continues, March or April could see 60d unique user counts over 2 million. Wicked growth! It’s probably a challenge to keep up.
The model of 60d active users isn’t perfect. I used fixed values for retention (85% loss after 30 days). In reality, retention rates will vary over time. It’s close enough (see comparison graph below) to do some assessments outside the range of data for which official numbers were provided.
A better approach might have been a linear regression on the known 60d active numbers.
Of course, having lots of people dropping by doesn’t do a company any good. But there are a couple of lessons here. In free-to-sample games, churn happens! The corollary is that a game based on non-traditional revenue streams (i.e. not subscriptions) needs to generate an incredible amount of traffic to gather solid users from whom to extract revenue.
“Hours of Usage” relative to the pool of active users has dropped. Early players were likely fanatical, on average, compared with the more recent crop. I would bet that translated into great conversion rates (from ‘free’ to ‘paying’ players) in the early days. Conversely, I’d expect conversion rates to be dropping.
For SL, where land-space (as I understand it) is tied to a server, increased hours of play might not translate directly to increased server costs (although, perhaps they pay for increased bandwidth). But it most certainly means increased server load, lag and therefore decreased player enjoyment.
I’ve read a lot recently about lag in SL. No doubt those users that do contribute to SL’s revenue feel the pain of the recent inrush.
Wikipedia lists SL concurrent users at a minimum of 17,000 and a maximum of 35,000 as of February 2007. The minimum doesn’t make sense if one simply averages usage hours over the number of hours in a month (in February 2007, the average concurrency would be about 16,400 by this method). Of course, I haven’t factored in any global downtime that might compress people into fewer hours.
The Second Life Business Research blog has a graph showing concurrency that seems to agree with Wikipedia, that concurrent users in February 2007 ranged between 15 and 35K, averaging at about 25,000.
I’m left wondering what the discrepancy is.
As with hours of play per 60d active, concurrency as a percentage of 60d is also dropping. Decreasing playtime and concurrency per 60d active user suggests that the community as a whole is becoming less invested in the game.
I would guess the cause to be the more casual player being attracted recently. I would also guess that more “casual players” means lower conversion.
If you compare the growth of subscription to the total growth of registered users, it’s recently been about 1%. Even the 60d active group only converts to subscriptions at a rate of 3 – 5% (down from mid-20’s in 2005).
In fairness, subscription is only one way to convert a user (SL has a number of other revenue streams beyond premium subscription).
Simply participating in the SL economy contributes to SL’s success even if it doesn’t put money in Linden Lab’s coffers: economic participation is one factor that contributes to the value of the Linden, thereby permitting Linden Labs to create (and sell) more of its virtual currency.
As with subscription conversion, economic participation also appears to be falling. Conversion by this measure has dropped from 27% of known 60d actives participating in the in game economy in September 2006 down to 17% in February 2007.
Even signing-in could be argued as having a net positive financial effect for Second Life. Real-world companies entering SL (CNET Network, Reuters, IBM) probably aren’t looking to make millions in Linden dollars: they’re looking for marketing opportunities (virtual billboard space, essentially).
Even non-economically-participating users who visit the virtual version of a real-world company have value. They increase the value of SL as a marketing vehicle for real world companies … this, in turn, could translate into more hard currency sales to real-world companies seeking virtual exposure.
Second Life is a more complicated business model, than, for example, Maple Story, which (as I understand it) focuses exclusively on micro-transactions directly with the ‘player’.
Whatever the approach to measuring conversion, it would appear to be falling. As one bloggist wrote (sorry, I forget who), SL appears to be a giant ‘try-me virus’ … attracting gawkers who have little intention or probability of contributing to the economy.
All in all, I wonder if it is a good thing for Linden Labs to focus so heavily on growth at the expense of player enjoyment (lag) or cost (more hardware / bandwidth).
They might be more financially successful to focus on factors that increase conversion: targeting ‘quality’ customers; reducing lag; adding game features; adding new revenue streams or improving ease of use.
The next step is to try to understand the revenue picture for SL.
SL draws one-time revenue from account set-up for alt’s, land auctions and island setups. On-going revenue is generated by the sale of the in-game currency, land usage fees (both mainland and island) and subscriptions. Anyone know of other sources?
For subscription revenue, I multiplied premium subscribers by $10. Many accounts will take advantage of prepaid cost savings. I didn’t bother to account for this. The subscription pricing FAQ is here.
Land pricing varies based on the size of the plot. I picked a number somewhere in the middle (see the chart). Linden provided total land area on their stats site. I assumed each subscriber gets 512m^2 free. The rest was calculated at the selected rate. Island pricing is $195 / month / region. Simple enough to calculate. In case you want to see, the pricing FAQS are here and here.
Information on $US value of in game currency sales by Linden Labs is available on the stats site.
A few other things I didn’t account for: one time set-up fees for alternate accounts ($10); one time set-up fees for islands; land auctions; discounts for charities and educational institutions.
As with all numbers presented here, this is simply an guess based on supposition, assumption and a few poorly understand usage figures released by Linden Labs.
Based on revenue one can then divide by active users to come up with an Average Revenue per Active User (ARAU). Somewhat different than the normal value of Average Revenue per Converted User … but as I argue, deciding which users are “converted” isn’t easy.
ARAU shows a decline from $3.45 in October to $2.14 in February. I wonder if I’ve missed a new revenue stream in SL’s evolving business model. Or, perhaps the rapid influx of people is detracting from SL economics … like gawkers at an accident impeding the flow of traffic.
Assuming ARAU of $2.71 (the average of Oct – Feb), and projecting backwards, you could arrive at a prediction of just over $8 million revenue in 2006. Using instead ARAU of $3.50 (basically, Oct 2006) and projecting backwards, you’d get about $9 million. This is really just a wild guess though. ARAU appears to have been higher in the early days.
I thought I could recall seeing some actual revenue figures for SL, but couldn’t manage to Google any. Anyone know if this is anywhere near reality?
In very round numbers, 10 skilled people (bennies, office, etc) is $1 million. So, roughly, $10 million in salary and sundry.
There’s still 2.5 hojillion servers and an electricity bill that surely rivals Star Gate Command.
As I understand it, each SIM (65536m^2) runs on one server. Based on total land stats, one can calculate how many servers they’re running. As for cost per server, I used server costs from Amazon’s Elastic Cloud as a rough estimate.
$0.10 per instance hour plus $0.20 per GB transfer, plus $0.15 per GB month storage. I made the assumption of 80 kbps down + 30 kbps up transfer based on this article and assumed 10GB storage per server.
Over 6 months ago, an CNET article suggested that Second Life operated 2,579 servers … back when they only had 240,000 registered users. I couldn’t find land sizes for that period to validate the server count calculations.
A more recent InformationWeek article says “Second Life runs on 2,000 Intel and Advanced Micro Devices servers in two co-location facilities in San Francisco and Dallas … Each geographic area in Second Life runs on a single instantiation of server software, called a simulator or ‘sim.’ And each sim runs on a separate processor on a server.”
My calculation suggests over 6,000 sims (therefore processors). I can only imagine that either the 2,000 server number is old, or each server is multi-processor, or that each facility runs 2,000 servers.
The article also says that SL maintains 34 terabytes of data. This would work out to about 5.5 GB per instance. I assumed 10.
I’m uncertain of how customer support is handled (whether that’s done by Second Life employees or outsourced). This might be an additional cost that should be factored in.
There’s really not enough information here to guess at bottom line. But, if we assume that Second Life has about 100 employees, that’s somewhere in the area of $10 million / yr ~ $850,000 / month. Add in $600,000 (in round numbers) in hosting costs for February 2007, and you have about $1,500,000 in costs (guestimated) for the month.
Contrast this with a revenue estimate of somewhere in the neighborhood of $3 million (guestimated) revenue for February 2007, and Linden is looking good.
A Business Week article from March 2006 says: “Even with plans to hire at least 30 more people to its existing staff of 70, Rosedale says the company is close to profitability. “Hopefully we’ll never need to raise private money again,” he says.”
If my analysis is anywhere near correct, it looks like SL has turned the corner.
SL might just do OK
They’re attracting the necessary foot traffic of potential payers and players.
Their business model is diverse: sales of islands to insane individuals and corporate entities looking for a virtual home; sales of mainland plots; sales of in game currency; and the odd subscription too.
The biggest challenge I would say they have is managing their wicked growth.
The biggest opportunity I see is improvement in either conversion rate or ARAU. Just imagine the revenue impact if they could do if they could double conversion.
This process has taught me a few things:
The mass market has a low retention rate, even when your online ‘game’ is free to play. Less than 15% of drive-by SL’ers stay active beyond a month.
Once retained as a player, the mass market has a very low conversion rate to paying player. Likely on the order of 3% or less. Multiply 15% retention by 3% conversion and the mass market delivers 4 or 5 paying players for every 1000 or so that bother to try out.
A tightly targeted audience (keen, committed ‘players’) can have a much higher conversion rate. Perhaps as high as 27%. A tightly targeted audience probably implies lower costs (fewer concurrent users to serve). I think the combination of high conversion and low costs is the smart move.
The old adage of 15%-20% concurrency doesn’t hold for casual free-to-play games. I could be as low as 1% for the mass market. For those keen, committed ‘players’ expect higher concurrency … perhaps in the range of 25%.
Two revenue streams are better than one. Provide customers with multiple ways to spend their money with you.
As I said earlier: I’d be much obliged of any wisdom readers might have regarding the business models of non-traditional multi-player games or worlds … likely retention rates; conversion rates; expected rates of concurrency.