ARPUs in social networks and social games: An acronym that needs scrutiny

How do
you monetize social networks and social games? Similar to the online games
space, there is a growing consensus that the answer is by complementing
advertising with a virtual goods micro-transaction business model. Social
networks and start-ups are positioning themselves accordingly. Hi5 has announced
its own virtual currency. Facebook is now implementing “credits” and MySpace seems
to be working on their own payments platform. Jambool provides virtual
currency, while Gambit, Zuora and Spare Change offer a variety of direct payment
methods. Users can pay for virtual goods by completing offers or surveys thanks
to SuperRewards, OfferPal and Sometrics while Zong and MobillCash let them pay
with their cellphones. Live Gamer provides a publisher-supported secondary
market for virtual good trading. It is only a matter of time until services
emerge that optimize electronic storefronts – layout, inventory, pricing and
promotions – and virtual goods marketing campaigns.

pitch for the micro-transaction business model is simple. There is a higher user
monetization relative to advertising revenues because of the type of engagement
that occurs in social communities and games. Central to this argument is the
metric for user monetization: Average Revenue per User (ARPU).The figure below shows several public
estimates of annual ARPUs[1].
The apparent wide discrepancy is misleading because of the different ways of
defining ARPU. If analyses based on ARPU segmentation, such as average ARPU
levels by game/app type and characteristics, or by social platform and region,
will inform key decisions about design and user acquisition marketing budgets,
these inconsistencies may result in wrong and costly decisions.

The R in ARPU

Bill Gurley estimates Facebook’s annual ARPU of $2.57[2] by
dividing the company’s annual revenue by its monthly active users. One might
mistakenly take this number as the ARPU of a Facebook user, yet this value does
not factor in that the user is also monetized by third party applications –
both through ads and micro-transactions. Developer Analytics estimated in August 2008
that more than 20 Facebook Apps had monetization potential of $2,000 per day[3].
Following the terms in its contract with application developers, Facebook
virtually sees none of this revenue. The only micro-transaction revenue that
Facebook itself generates is through its virtual gifts, which accounts for
$50-$60MM[4] of
annual revenue. In other words, $2.57 is how much Facebook, as the platform,
monetizes on an average user, not the monetization value of a Facebook user. If
Facebook wanted to increase its own ARPU, it could simply develop more of its
own apps with micro-transaction models or it could change terms of contracts
with developers in order to get a larger cut from their revenue.

Liew from Lightspeed Ventures estimates ARPU for Jagex’s MMO Runescape
(freemium game) as $0.84/month/user[5].
In contrast, all user monetization in Runescape hits Jagex’s top line as they have
no third party apps.

Why are
there no public estimates on the monetization value of a Facebook user? It is
hard to get data to estimate the aggregate revenue across all Facebook app developers.

The U in ARPU

At the
SXSW conference, Susan Choe, CEO of Outspark (a publisher of Free-to-Play online games),
announced $50 monthly ARPUs, creating strong reactions until the audience
started concluding that she was probably referring to ARPPU – average revenue per
paying user. This illustrates the definition
issue with “users”. Different definitions of a user dramatically skew the ARPU numbers:

Registered users: They sign up or install
game/app, but may use it very infrequently (if at all) and they never pay.
Minimal or non-existent monetization.

Active users: They use app or play game with a
certain frequency, but never pay for anything. Monetization only through ad

Paying users: They pay for purchasing digital
assets or for subscriptions at least once or with a certain frequency.

within these categories, companies might define users differently. The
differences in the numbers because of varying definitions are in the orders of
magnitude. As an example, in his ARPU vs.
blog post[6],
Raph Koster estimates monthly ARPUs in freemium successful models as
$0.50-$1.50, versus $30-$35 monthly ARPPUs for subscription games.

to Jeremy Liew’s Runescape ARPU estimate, he reached $0.84/month/user by
dividing the monthly runrate ($5MM) by the 6MM Runescape players, 5MM of which
play for free. If he had used, the 1MM subscribing paying users, he would have concluded
with an ARPU of $5.00/month/user.

Proposed Standard Definition

Many of
these inconsistencies can be addressed if as an industry we were to adopt a
standard definition. Here are my thoughts around a standard definition but
would like to hear yours.

Average Revenue per User where:

Revenue is defined to include all monetization
of the user within the social networking platform, virtual world or game,

All revenue models: advertising,
micro-transactions, subscriptions, player-to-player, etc. This introduces a
“share-of-wallet” mentality where the revenue model ARPU can be specified (e.g.
advertising ARPU, micro-transaction ARPU)

All parties that monetize on the user within a
platform, world or game. Hence, the ARPU of a Facebook user includes both
revenues to the Facebook platform and to its third-party application

User is defined as a registered user – a user
who has signed up or installed software

Part of the ARPU optimization process will
include the conversion of registered users into active ones, and in turn, into
paying ones

Time horizon is specified (e.g. monthly ARPU
versus annual ARPU)

Average Revenue per Paying User same as ARPU, but where:

Paying User is defined as any registered user
who made any payment since signing up, whether for a purchase or a

The process of payment will have required the
user to set up a payment method (e.g. input credit card details, link to Paypal
account). Hence, the barrier for further payments is significantly lower.

proposed definition includes several key caveats:

Differences in registration process (e.g.
filling in long list of user details versus auto-registration by visiting a
site) will still make apple-to-apple comparisons slightly misleading

At times, subscription and micro-transactions
are not fully separated. In some games and worlds, paying a subscription also
gets you a certain amount of virtual currency for micro-transactions

Difficulty of ARPU measurement across several
parties (e.g. a platform and hundreds of application developers) and across
revenue models

ARPU definition will require more specificity,
e.g. monthly micro-transaction ARPU

As the
micro-transaction business model gets more traction, we will see a spike in
claims of ARPU figures. These will vary wildly – at times because of their
nature, at times because of the metric definition. Caveat emptor.

About the author

Pourshasb is director of corporate development at Live Gamer. His background
includes Business Development at MTV Networks, where he led M&A deals with
social media properties, and a Senior Project Manager role in strategy
consulting at Oliver Wyman, specializing in Emerging Markets. He holds an MBA
from Harvard Business School,
during which he worked part-time at venture capital firm General Catalyst,
helping launch an online advertising company. Other work experience includes Goldman
Sachs and PWC, and he also holds a Masters in Engineering from Imperial College,

Gamer is leading the way in legitimizing player-to-player virtual economies,
working in partnership with publishers and developers to realize the growth and
impact RMT of virtual goods has on game design, player experience and online

Sources for the data points are listed in the following footnotes.

Above the Crowd, How to Monetize a Social Network, 3/9/09

Raph Koster, Metaplace, 3/16/08,


4 thoughts on “ARPUs in social networks and social games: An acronym that needs scrutiny

  1. Insightful, well researched and thought provoking article. Your credentials and industry experience also speak for themself. However I disagree on the broad definition of revenue as inclusive of 3rd party apps etc. The platform ARPU should refer only to the revenue actually received by the platform otherwise comparisons become very hard.

    For example if facebook hosts a legal gambling application that application would probably have pretty huge ARPU but thats a factor of my wanting to gamble and that revenue wouldn’t really be available to facebook as a social networking platform to capture. Also the line becomes pretty blurry – would you consider my buying an item on Craigslist from another user a “player to player” transaction to include in ARPU – probably not, but buying something in WOW would qualify.

    Yours truly

  2. Thank you for the article Nima.

    Samir presented critique on the proposed model’s revenue part. This was something that I would have highlighted as well. I think he made the point well and gave illustrating examples, thus I will not get into that, but:

    The definition of user to include only registered users might not take into account all the relevant users in all cases, although the restriction of registration draws a clear line. Disregarding the registration as criteria clearly makes it more difficult to estimate the size of user base. Still, in many services majority of users will not register at all, although purchasing virtual assets commonly requires some kind of registration. But even this is not always the case. Particularly with services that have advertising lead revenue generation, the user base commonly consist of non-registered users, who still conduct revenue generating behaviour within the service. This problem touches both the definition of users in presented proposition of defining ARPU and ARPPU.

    One has to also keep in mind that the ARPU is not actually a very good metric for comparing revenues of services. ARPU is highly service nature dependent, whereas services harnessing micro-transactions are usually looking at lower ARPU than e.g. subscription based services, they can nevertheless be generating much more revenue, due to the large user bases. There is also lot of variation in costs, service type, etc. ARPU measures only how effectively service monetizes its users and does not hint much about the actual profits that are being made. Other metrics such as ARP registered users, paying users and active users are other more specific tools in analysing how a service is monetizing different segments. It is service specific what other such metrics are deemed necessary and useful. For example, a firm selling virtual assets might be interested in how avatar class, let’s say, hunters are monetized, i.e. how well the game incentives this segment to purchase virtual assets. That all being said, ARPU seems to be more of an inner-organizational metrics for measuring own monetization effectiveness and less suitable for comparing between different services.

  3. Thanks for the thought-provoking article, Nima.

    I think both the article as well as the comments above show that while ARPU and related concepts need to be scrutinised and criticised, it’s more difficult to come up with acceptable alternatives. The passage cited below made me think that perhaps the key to defining a useful metric is to first articulate very precisely what the metric is intended to be used for:

    If analyses based on ARPU segmentation, such as average ARPU levels by game/app type and characteristics, or by social platform and region, will inform key decisions about design and user acquisition marketing budgets, these inconsistencies may result in wrong and costly decisions.

    Thus, if you operate a service and want to know what segment to spend your marketing budget on, you only need an internally consistent metric to compare the segments’ revenue potential, as Juho suggested. On the other hand, on a different day you might want to compare your service to other services in the same category (substitutes) to see whether your monetisation capability is on par with them. For this you would use a differently defined metric. But if you’re a venture capitalist trying to decide whether to fund a social gaming startup or a mobile social networking app, yet another metric is probably needed.

    When people blog about the ARPU of such-and-such service or talk about it at conferences, they are probably implicitly using it in the latter two senses: 1) to argue that service X has a better monetisation rate compared to other similar services with similar content (and similar costs?), suggesting that those other services could potentially reach similar ARPUs if they improved their monetisation; and 2) to argue that industry X is a better investment than industry Y — for example, that social gaming is generating more buck for bang (however bang is defined) than virtual worlds.

    I would be interested to see more stuff about this on VERN. I would also be interested to hear what other uses people put these metrics into.

  4. Very interesting read…my first mmo game was Ultima Online, with a strictly p2p model per month, but as I have delved into other games I noticed I have spent more money on the “free” games with micro transactions within them. I guess it’s the same principle as “death by a thousand cuts” only rich by a million transactions. I think it’s a slippery slope though to create your own currency, unless you are a huge powerhouse such as fb, microsoft, or Hi5. The social aspect is there for networks too. Sites like are popping up to cater to gamers so it would only be a natural step to have some type of virtual currency on networks where people are already “used” to it, or less jaded towards the idea.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s