The social and mobile gaming industry is a sector with strong growth that attracts more and more established companies as well as start-up firms. Reliable sales forecasts are necessary to find investors. However, forecasting in a rapid growth industry like gaming can be a tricky affair.
Classically, there are two approaches to revenue forecasting: top-down and bottom-up.
- Top-down: The top-down approach usually starts with an estimation of the market size and goals for the market share that is to be achieved. Top-down forecasting is rather suitable for developed, non-volatile markets like consumer goods or cars. For social or mobile gaming, estimating (or even predicting) the market size, perhaps even split by genre, can become a strenous exercise, as the market is still evolving.
- Bottom-up: Using the bottom-up approach, you need to consider all the constituents of a company's revenues, and estimate them. You should try to estimate the numbers as low-level as possible. The typical constituents would be the number of downloads for the different platforms, the number of items sold for each type of in-game virtual goods, and the click-rate and cost-per-click (CPC) for in-game advertising.
Some practitioners argue that, in many cases with financial modelling, the bottom-up approach leads to a false feeling of accuracy: The model incorporates a lot of variables, mingles them together somehow and presents a result that is seemingly reliable and supposedly much better than just roughly estimating "sales will grow by 10 to 15% next year". Whether that is true or not, the most recommended approach in practice is to start out with the bottom-up approach, and then optionally verify the plausibility of the result with a top-down estimate. There are two good reasons to favour the bottom-up approach:
- Firstly, while the forecast may or may not be correct at all, the separation of input variables in a detailed bottom-up forecasting model allows for much greater flexibility. You could easily include a sensitivity analysis that would show how the results would change if one or more input variables would change. For example, how would your cash flow develop if the user base would grow at the same speed, but the players would spend 20% less than expected on virtual goods? The numbers may never be completely accurate, but the user gets a good idea of the impact of certain events.
- Secondly, even if the feeling of accuracy and reliability is unjust - having a seemingly reliable business plan is very important for small and growing companies that are looking for an investor. The more sophisticated your forecasts are, the more you show that you understand your industry, that you have thought about the business model, and that you have considered every possible way to increase revenues.
Forecasting the user base
Whatever the most important revenue source for the gaming company in question, the key input to all forecasts will always be the user base. Download revenues, virtual goods sales and in-game advertising all depend on how many players you can attract. For social and mobile gaming companies, the user base is normally easy to measure because people need to sign up online and/or download the game through an app store.
In order to make the forecast as transparent as possible, you should rely on the concept of control accounts. Every step in the forecast of the number of users should be separate and easily explained. In the most simple forecast, you should include 4 positions:
- User base - beginning of the period
- New users
- User loss
- User base - end of the period
You start out with the user base at the beginning of the period (or "bop"), add all new users, subtract all lost users and get an estimate of the user base at the end of the period (or "eop"); this value will obviously also be the user base at the beginning of the next period.
Of course, you do not have to limit your financial model to just one row called "new users". You could split this up to make effects more transparent. For example, you may have one row for regular app sales, and one for "campaign-related spikes", e.g. if you pay money to a website to be featured or something similar.
The user loss is a typical phenomenon: After a while, people will get bored by a game and delete their account or delete the app, respectively. Whether to predict this as a percentage of players lost or as a percentage of players retained (the user retention rate) is rather a matter of personal preference. Either way, the figure is hard to predict and requires some industry expertise, as well as an understanding of different genres. Another way to think about this is to ask yourself how long you think the average user will be active in the game. If you think 12 months seems realistic, a monthly user loss rate of 1/12=8.3% (or roughly 8%) may be a good estimate for your business plan. You should also consider that a number of users might delete the game / their account right away if they do not like it (especially if it was free to download/register) or if they face compatibility issues.
Forecasting user activity
The revenues that a game generates not only depend on the number of users, but also their activity. In the app and gaming world, there are two key measures to indicate activity for a game: daily active users (DAU) and monthly active users (MAU). Both measures indicate how many unique users are active on a given day or in a given month. When dividing a month's average DAU by that month's MAU, you get the engagement ratio (DAU/MAU). It shows how many of your active players are active on a given day. For mobile games, it is very difficult to get reliable DAU and MAU data. For social games, however, appdata.com is a great source. Looking for benchmark engagement ratios, you can see that successful companies like Zynga, Electronic Arts and wooga realise DAU/MAU values of 20 to 25% for their Facebook games.
Revenue sources for a social and/or mobile game
Sales of the game itself
This primarily applies to mobile game developers - and it's as simple as can be:
revenues per period = price per download × new users per period
The "price per download" will not be what the user pays to the app store, but rather what your firm receives after all app-store-related costs.
Quite obviously, the number of new users here should be linked to the one used in the control account for the same period - see above.
Virtual goods and premium content
The so-called "freemium" model is very popular among social and mobile games: Players can play the basic game for free, but have to pay for virtual goods (such as in-game currency or items) or premium content (such as additional levels) - hence the term "free-mium". The most common ratios in this context are ARPU (average revenue per user) and ARPPU (average revenue per paying user).
virtual goods revenue = ARPU × MAU
virtual goods revenue = ARPPU × paying users
Prior to a game's launch, ARPU and ARPPU should be estimated based on similar games (as long as data is available), and a logical assessment of the characteristics of the expected player demographics: 16-20 year old players in high-wage countries might be more likely to spend money on virtual goods than 10-12 year old players in low-wage countries. (Remember that the purpose of this detailed way of modelling is not to ensure the numbers are 100% correct. Rather, this ensures that all business plan assumptions are plausible, and provides a framework to see the effects of changes in the inputs.)
In-game advertising by third parties
Again, this primarily applies to mobile game developers. Social games that run via Facebook or Google+ can normally not show their own advertising because the social network will want to make money off advertising itself. However, online games that run independently (such as the classic CSManager) usually do show ads.
The forecast of the ad revenues, as many other ratios explained in this article, uses inputs that may be difficult to estimate at first - but will be much easier to estimate once you have acquired some experience and have some historical data to rely on:
ad revenues per month = ad impressions per month × CTR × CPC
...with CTR being the click-through rate, meaning the percentage of users that are likely to click an ad, and CPC being the cost per click, i.e. the money that you receive (on average) for each click on an ad. How well you can predict the number of ad impressions depends on how detailed you can mine the data of your players' behaviour. You could tackle the issue like this:
ad impressions per month = MAU × average time spent in game per month
Other revenue sources
Successful games can sometimes tap additional revenue sources that may be difficult to forecast at the very beginning. For example, it is not unusual to grant licences to third parties for ports to other platforms or markets. Another common practice is to sell or licence parts of the code (the "engine") to another developer for a similar game with different content design. This is similar to the way in which some classical video game developers have improved their revenues, most notably id Software.
In very rare cases, if a social or mobile game is extremely successful, there may be revenues from merchandising products - think of t-shirts, or even the plush animals inspired by Rovio's Angry Birds.
The example above assumes an average user activity, average conversion rates and average spending behaviour across the whole user base. This, of course, is a simplification. There are almost always different segments of users, ranging from the casual gamer who barely spends the tiniest amount of money to the hard-core player who will spend time and money almost every day.
These player segments can be split and shown separately, with more specific user loss/retention rates and of course ARPU values. If you have historic data available for the game you are analysing, or older games you produced, you may be able to forecast the shares of the player segments to a certain degree - and you may be able to forecast impacts of certain changes to the game's rules.
For example, assume that you would like to analyse potential effects of introducing activity bonuses in the game. Some games do this explicitly (by having something like "activity scores" based on the number of logins), and many more games do this implicitly (by granting rewards for perfect timing over hour- or day-long timeframes, such as in CityVille). An Excel model could reflect manifold effects of rule changes:
- Increased DAU, because a lot of players will play more often
- Increased ARPU, because players might spend more time in the game, perhaps spending more money on in-game goods and clicking on ads more often.
- Decreased MAU, because casual players will be more strongly penalised and therefore be less motivated.
- Increased DAU/MAU, because the DAU would rise and the MAU would fall.
Modelling and presenting scenarios
In the context of a business plan, and especially when the business plan is used to look for external investors, it is common to readily provide a few sensitivies in the forecasts of the financial model. The most common way is to present one base case that is the "original" forecast, a best case that shows how the company or project would evolve if sales end up much better than expected, and a worst case that shows the consequences of sales numbers that are much worse than expected. A few rules of thumb:
- Having 3 general cases in a first business plan is the most common way. You can provide up to 5 if those make sense, but for more than that there would need to be a very specific reason.
- Do not necessarily make your "worst case" as bad as it could possibly turn out. Make it seem decent somehow, but still bad enough to be credible as a "worst case". Experienced investors will naturally assume that something labelled "worst case" is not as bad as it can get and implicitly think of an even worse scenario.
- The financials you present at first should include all the relevant positions in detail, and sum up all the irrelevant positions as "other items". If an investor is still interested after having seen the high-level financials, they can look at the detailed outputs of your financial model.