[Editor’s note: This article was written by Bill Gurley, a venture capitalist at Benchmark Capital. He published it on his personal blog, abovethecrowd.com, last week. Accounting for virtual goods is becoming a bigger industry issue, so we’re running the article here with his permission.]
American journalists and corporate executives have been slow to appreciate the beauty, brilliance, and consumer allure of the virtual goods business model.It’s not that they did not have data points – China is chock full of multi-US$billion market capitalization companies that are based on this business model. That said, many luddites predicted it was an “Eastern” fascination that would never spread to the West. They never fully understood it.
As a result of this headstrong denial, I have often wondered what data point would finally convince me that the West had fully accepted the reality of the virtual goods business model. Last week I received my answer. Jeff Grabow from Ernst and Young asked my partner Mitch Lasky and I to sit down with Mick Bobroff, an audit partner developing an expertise in virtual item based revenue recognition. Now I wasn’t exactly waiting for a sign from God or anything – rather just a small signal that confirmed this new model was legit. Having an audit partner at a top three accounting firm become an expert certainly qualifies as a step in the right direction.
Mick had prepared a remarkably succinct and information rich presentation (they are working on a white paper I can post later). I was fairly excited to go through it – at least as excited as anyone should get when discussing accounting principles. Here is a summary of what Mick had to say in his presentation titled “Revenue Recognition Considerations for the Sale of Virtual Goods”. [If you want to reach Mick, his contact info: 415-894-8205, email@example.com]
- [Legal Point First] Michael made it clear that this document represented general observations and should not be used specifically as accounting advice. I understand and concur with regards to this post also. For your own books verify with your own accountant.
- There are already a ton of companies that trade on American stock exchanges (NYSE, NASDAQ) that use virtual goods models and adhere to GAAP. To the point above, they are all in China. Examples: ChangYou (CYOU), Giant (GA), NetEase (NTES), Perfect World (PWRD), The 9 (NCTY), Shanda Games (GAME). [For the record, TenCent is Hong Kong listed.] The current GAAP revenue recognition policies were honed with these companies.
- When a company sells virtual currency, this is not a revenue event (even though it may clearly be a cash event). When purchased but not yet used, virtual currency sits on the balance sheet as a customer deposit or deferred revenue (i.e. a liability).
- Revenue recognition commences when virtual goods are bought with virtual currency by the consumer. Exactly how it “commences” depends on the following.
- There are two categories of virtual goods – (1) consumable items that are used once and gone, and (2) durable items that “work” over an extended period of time.
- For “consumable items” you can recognize revenue when it is consumed.
- For durable items (which many are), things are much trickier. You need to amortize the revenue (linearly) over the useful life of the good, or the average life of the actual user (i.e. – what is the average customer life of your customer?). This is a messy problem, especially when you understand how difficult it is to measure “customer life” when some customers never pay and others come and go in fairly random patterns. Also, your “average customer life” may change over time creating very complex accruals.
- The bottom line: getting this right requires quite a few database entries for tracking the sale and usage of every single virtual good sold in your digital world, in addition to the supply and usage of each virtual currency account, and the activity levels of each user (to estimate average life).
These policies were not particularly surprising. That said, when I was listening to the complications of the “durable item” revenue accounting, it reminded me of something I learned for the early leaders in the virtual items space — innovative Korean companies such as Nexon.
The “Rental” Model
About four to five years ago, the team at LindenLab (SecondLife) held a pizza night at their offices with the goal of learning more about the virtual item games that were wildly popular in Korea. We invited two bilingual gamers to install and play Audition, Kartrider, and FreeStyle. My big takeaway from that night was that not one of these titles actual allowed for the “sale” of virtual goods. Rather, each virtual item could only be “rented.” In each case, the user was given the option of one, seven, or thirty-day rental. I assumed this was Darwinian, and immediately began to wonder why “renting” might be better than outright ownership when it comes to virtual goods.
- In world inventory gluts. As virtual worlds mature, they often suffer from game-wide inventory glut. Items that were once useful to newbies become throw-aways for the more advanced user, and can literally pollute the world and compromise the in-world economy. Allowing rental is like having free garbage collection. Everything eventually goes away.
- User inventory clutter. More advanced users typically have a huge problem managing large inventories of items. Also, many items are trend-oriented and trends change. With the rental model, no user sits around thinking “wow, why did I really buy that two months ago and what do I do with it now?,” and “why am I buying all this stuff?” The rental model simplifies inventory management for the user.
- More marketing opportunities. When an item expires, it offers a unique time to re-market to the user for either an extension of the current good, to a trade to a newer, fresher, and perhaps more interesting item.
- Price segmentation. By offering 1, 7, and 30 day rentals, the merchant has basically price-segmented the market. This theoretically allows more users to experience the good than may have with a single, and arguably higher, price point.
- Good business. Why sell something that lasts forever if you can sell something that has to be naturally repurchased?
- Simpler accounting. I didn’t think of this sixth point until my meeting last week. The rental model does away with the notion of a “durable” virtual good, as they all expire. What’s more the time frame over which you recognize the revenue is now fixed at 1, 7, or 30 days. This dramatically reduces the accounting complexity.
Thanks again to Michael and Jeff at E&Y for reaching out and setting up the meeting. It’s great to recognize that virtual goods businesses are finally mainstream here in North America, and that they even have their own appropriate accounting policies. I also appreciate having one more reason to favor “rentals” vs “sales” when it comes to virtual items.
[I have received several comments that concern this post and how it relates to SecondLife. For those of you that don’t know, SecondLife doesn’t actually sell virtual items, its residents do. As such, this post does not relate to SecondLife at all. It pertains to the 98% of virtual worlds where the hosting companies ALSO is in the digital goods business. Nothing would stop SL from offering rental as a choice to its developers, but the main message is that this post does not pertain to SL (which has a different business model altogether.)]