Summary: An analysis of QQ.com – a profitable Chinese social networking and instant messaging service with 1 billion usernames, 75 million peak concurrent users, and plans to grow beyond China.
Introduction
The world is full of fast-growing, hyper-fashionable social networking and user-generated content plays. Almost to a man, they lack one thing – profits, or even revenues. An English-speaking technology media and analyst/investor community obsessed by the US West Coast has practically ignored QQ.com, one example of spectacular success, because it’s Chinese.
A Profitable and Valuable Social Network
At the 30th of June, Tencent (QQ’s owners) had thrown off RMB993 million (US$145 million) in free cash in six months, even after spending RMB1.9bn in CAPEX and a further RMB593 million in financing costs. For comparison, Facebook went marginally cashflow positive for the first time in August and isn’t yet profitable.
The bottom line is impressive too; at the last count, Tencent’s gross margin was at 67.3% and net margin was 41.75% – this smashes HP’s investment criterion of “fascinating margins”, i.e. 45% gross, and Iliad’s 70% ROI on new fibre deployment. We previously estimated the gross margin for October 2008 as 63.5%, so it appears that things have consistently been going well for QQ.
The shares (listed in Hong Kong) have gone from HK$60 to 120 since April, showing that this performance is also attracting plenty of demand from investors – albeit at a somewhat toppy price/earnings ratio of over 50.
Nearly a Billion ‘Users’
There were 990 million user identities on QQ as at the 30th of June, 2009. Given the current growth rate, the billionth user will almost certainly be announced in the next quarterly results – but a nontrivial percentage of these are inactive, are multiple aliases, or are spambots. [NB. This is true of all IM communities except, perhaps, for the 17 million users of IBM Lotus Notes Sametime inside their enterprise firewalls, as we pointed out in the Consumer Voice & Messaging 2.0 strategy report.]
As impressive as this is, instant messaging user bases are usually only weakly bound to the service, they are usually non-paying, and many people have multiple usernames. A more useful metric is peak concurrent users – the maximum number of users simultaneously logged in during the period in question. To be counted, a user name has to be active in that they are online, so it’s reasonable to deduce that they exist. It doesn’t prove that they are a human being (or for that matter a useful application rather than a pest); however, whether or not a logged-in user is human, they are consuming system resources.
So, measuring peak concurrent users provides us both with better data on uptake and a more useful indicator of capacity related costs. It’s a standard telecommunications engineering principle to “provision for the peak” – that is to say, it’s useless to build a network with only sufficient capacity for the average traffic, as 50% of the time it will be congested and probably non-functional through overload. To be available, the system must supply enough spare capacity to handle the peaks in demand. Peak load determines scale, and hence cost.
In 2008, at various times, QQ’s parent company Tencent claimed to have between 355 and 570 million users. At the end of June, 2009, the user count stood at 990 million – so the nominal user base had roughly doubled. In 2008, peak concurrent users were 45.3 million, growing to 65 million in June 2009. According to QQ.com’s live statistics readout (you can watch it grow in real time here), the record at time of writing was 79 million. According to Alexa, 3.26% of global Web users visited one of the various qq.com sites in September 2009.

For comparison, Skype’s all-time peak concurrent user count is 15 million, although it has the advantage of using user-provided infrastructure, whereas QQ has a client-server architecture and therefore a constant need for rack-space.
Not just users, but Paying Users
In 2007, out of 12 million peak concurrent users, 7.3 million had spent money with QQ, or to put it another way, 61% of verifiable QQ users were buying value-added services. (How many mobile operators can claim that?)
In March, 2009, we thought it unlikely that this high proportion would continue to pay as the service grew – and that it was quite possible that the 7.3 million earlier payers were dominated by early adopters and power users, so that future recruits would be less committed to the community, less geeky, and lower-income.
However, when Tencent’s Q1 results appeared at the end of March, 36.9 million users had purchased value-added services during the quarter, growing at a monthly rate of 8.4% to reach 40 million by the end of June. This latter figure was against a concurrent user base of 65 million, meaning that 62% of concurrent users were paying users.
We think this is an impressively high proportion at such volumes, and suggests that the revenue may scale reasonably well as it grows penetration further. As one might expect the cost model of such a volume business to scale efficiently, this implies further prospects of profitability. It is likely that such thoughts are one of the influences on the aforementioned growth in QQ’s share valuation.
So, how did they do it?

In our Serving the Digital Generation Strategy Report, we identified a list of key factors that anyone who wants to attract the customer of the future would have to address, which together describe what we call the participation imperative. Specifically, four axes define the customer’s aims:
- To interact socially with a peer group
- To personalise and customise their environment
- To express creativity – e.g. user generated content
- To maintain privacy/anonymity or seek notoriety
These require and depend upon four key affordances:
- Portability – broad ability to work across multiple PCs, mobiles
- Payments – virtual currencies, transactions
- Feedback – ratings, comments, discussion, personalisation, hackable APIs
- A directory – to find other people
We assess that QQ hits 7 out of 8 criteria squarely. Really, the only one they don’t cover is privacy – although they do have rich presence-and-availability control, it’s in the nature of such a community that going offline could be a noticeable act, and there have been problems with the Public Security Bureau (Chinese secret police).
NB. The Customer of the Future can be a complex and powerful character. When the Shanghai PSB demanded that QQ filter references to the Diayou islands (a controversial nationalist cause in China), the ensuing user revolt caused even the PSB to back off.]
QQ caters to user creativity and the need for personalisation much more deeply than most social networks with the possible exception of Facebook. Although officially proprietary, the system API is documented and QQ, the company, positively encourages a hacker ecosystem of interesting new applications. This goes some way beyond the skins and avatars most socnets offer. Similarly, you can’t offer more effective feedback to more advanced users than the ability to tinker with the works. Portability is well catered for – there are multiple client applications, SMS integration, various mobile clients, and the Web site.
Print your own Digital Money
QQ’s in-world digital currency is no trivial add-on. QQ derives revenue from selling applications, other in-game goods, and extra services such as a blog, games, and a streaming music service, in return for its internal digital currency. This market creates a sink for the digital currency, and therefore gives it value, which creates a further demand for it as a gift and reputation good. It shares revenue from the store with the creators of in-game goods, thus feeding user creativity.
In Telco 2.0 terms, QQ’s business model is collecting money from the downstream side and subsidising the upstream partners, in order to encourage the creation of saleable goods and the purchase of digital currency. In return for their participation, users get the core functions of the directory and the messaging layer to service their peer group and burnish their on-line identity.
In-World Currency dwarfs Advertising
Although QQ also does contextual advertising, its core business is the in-world economy. We remarked back in March that the ad business was overshadowed by the VAS business, and this is even more true now. Online advertising grew just under 10% year-on-year, but now makes up just 9% of total revenues, falling from 11%. Internet VAS revenues were up 107% and mobile VAS was up 38%.
In part, this is the unavoidable downside of being hackable; advertising is a tax on your attention, so some people will want to be rid of it. Just as many Mozilla Firefox users install Adblock Plus to screen out Web advertising, multiple unofficial QQ clients exist that strip the ads. But if the users buy the clients from the QQ Store, who’s complaining?
QQ’s ‘Two-Sided’ Business Model Strategy
We’ve identified three types of generic ‘two-sided’ business model strategy, and concluded that the most successful companies were those who operated at the creative edge between each type.
- Strategy One involves giving away services before and after a transaction, and collecting a percentage of the transaction. Think Amazon – or a casino.
- Strategy Two involves giving something away to create a trading hub, then selling something to the crowd. Think of the original Lloyds’ Coffee House – it didn’t write marine insurance itself, it sold coffee to the insurance brokers, who came for the liquidity and rumours, and stayed for the coffee.
- Strategy Three involves selling access for third parties to the trading hub – like BAA plc renting shops at Heathrow Airport, or Google giving away a whole range of services in order to create inventory it can sell adverts next to.
QQ would initially appear to straddle Strategies Two (selling to the crowd) and Three (charging for access) in the two-sided business model. But the domination of in-world trade over advertising in its P&L statement suggests something else – much of what it sells to the crowd originates in the crowd. Isn’t this an example of the Amazon-like Strategy One, facilitating transactions in return for a turn on the deal? If so, they’ve brought off the impressive feat of exploiting creative ambiguity between all three.
Next: your market?
Where does QQ go from here? The answer appears to be “right here” – in August 2009, Tencent launched an English-language portal (imqq.com). Interestingly, the site is marketed directly at business, which is an extension of a strategy shift they have already undertaken in China. For some time, Tencent has been marketing a version of the client at business users which borrows the look-and-feel of Microsoft Live Messenger (apparently being boring can be a valid strategy).
The business version of QQ is paid for – sensibly in our view, Tencent don’t expect small companies to be spending much time trying to achieve legendary status in the QQ user community. As (supposed) serious, responsible adults, they’re meant to have a secure identity and reputation already, so they’re not likely to contribute that much to the in-world economy by trying to burnish them. Therefore, a traditional, one-sided model is being used to derive revenue from this submarket.
Conclusion: Watch with Care
Our conclusion is at this stage that the Telcos who aren’t yet familiar with QQ should keep a close watch on them in both home and away markets. At a minimum there’s a lot to be learned from how this smart and complex operator employs the ‘two-sided’ business models. At other extremes are competitor threat and partner opportunity scenarios that we’ll be looking at in more depth in our future analysis.
Even though there are a lot of mobile industry execs with scars from trying to transplant successes from (usually) Japan into WENA (Western Europe & North American) markets, complacency would be extremely unwise faced with a potential competitor that has demonstrated such a deft grasp of two-sided business models, such a close understanding of user needs, and such a solid base of competence in high scalability Internet engineering.
And Finally…
Bill Gates recently gave a speech in which he claimed that two out of the five most profitable firms in China “don’t pay for their software”. He was telling the truth, in a sense; a quick “curl -i im.qq.com” demonstrates that Tencent isn’t paying a penny for its server software – the site is served with Apache running on BSD Unix machines. That may not be what Bill meant, but perhaps he should have.