Getting Started With GameAnalytics

Analytics is a powerful tool for informing decision making at strategic and operational levels in game development. It will help you answer key questions about design, fun, engagement, play, monetization and more. But it is a relatively new process in game development and there is not a lot of knowledge available out there for the non-expert. In this blog post, we will try to help you get started on analytics.

Analytics for games has become a much-discussed topic in game development in recent years, not only for F2P but across the entire industry. However, being a somewhat new phenomenon there is not a lot of easily accessible knowledge available for the non-expert, and what knowledge is available is highly unstructured. Simply Googling different analytics terms leads to an array of information about analytics as it applies across a range of industrial sectors, which is not terribly useful. Siphoning away the stuff that deals not with games but e.g. healthcare, stock trading or web store customer analysis, and focusing on games leaves a number of blog posts, a few good Gamasutra and Game Developer Magazine articles, presentations from industry events like the Game Developers Conference and Casual Connect, research articles, white papers and documentation from a variety of third-party analytics providers like our own Game Analytics. These various resources in turn introduce a huge range of acronyms like DAU, MAU, ARRPU and LTV, which can be used to mean different things in different contexts. There is also a heavy emphasis on analytics used for monetization, i.e. the process of analyzing player behavior towards maximizing revenue, and on Free-to-Play/online games – with much less knowledge available online about using analytics to inform design towards increasing fun and engagement. In other words, the knowledge is fragmented and has a lot of uncovered but vital areas to those just starting out on the path to integrating analytics in their games.

The goal of this post and those following it in the series, is to try to collate some of the most pertinent available knowledge for the non-expert about how to get started with analytics, irrespective of the game type or the delivery platform. The post by no means covers everything you need to know as a game development company to reach mature business intelligence practices, but we will point the way and suggest where to read more about the various topics involved.

Getting started: A 12 step model

There are many ways to get started with analytics, many of which are good. The model we will discuss here is fundamentally based on the standard model for Knowledge Discovery used in business intelligence virtually everywhere, but adapted to the specific situation of guiding a non-expert in game analytics towards implementing and gaining insights from analytics practices. We are here assuming no prior knowledge of analytics, and will take the time to identify some of the key terms. If you already have experience with analytics work, other material on this blog should be of interest.

As with everything else worth doing, game analytics takes some time to learn how to do right. There are definitely quick ways to get started, but it pays off spending a few hours reading, thinking, planning and testing before embarking on integrating data-driven practices in game development.

We would have liked to find a cool acronym for the model, but alas could not – suggestions are welcome – there might even be a small price!

The model can be summarized in the following steps:

1)     Basics: The first step is to form an understanding about what analytics is and the specific role analytics play in game development and –management. Analytics is much more than monetization – it is a tool for the entire company.

2)     Key terminology: The second step is to learn what the key terms in game analytics are and what they actually mean. Due to the newness of the use of analytics in games, there is some disparity in how essential terms are employed. Without understanding key terms, looking for knowledge is seriously hampered so it is worth getting the facts right from the onset.

3)     Read: The second step is to browse and peruse some of the key material currently available. This may seem like a strange second step but it is nonetheless our experience that it is vital to do a bit of exploration about what other people are doing in different contexts before even considering what do to in your own game.

4)     Process: The third step is to understand the analytics process from the point where questions are asked to when they are answered and game designs or company practices changed as a result. There are a number of key considerations impacting the analytics process, notably the specific requirements of the stakeholders involved.

5)     Goals: Knowing what game analytics is and having an idea about how it is being used around the game industry is, the next step is to develop goals for what you want to do with your game/company and analytics. Because analytics is such a broad area, that can be used at strategic and operational levels, brainstorming about the full spectrum of potential uses and goals is a great first step, but should be followed by a tight focusing on a small subset of goals for initial implementation and testing. This allows for an evolutionary prototyping strategy where analytics is matured in an iterative fashion, and minimizes the starting investment. Setting goals can be incredibly hard to begin with, especially in the earliest stages of a design process, so iterative development helps making analytics integration a learning experience, which is highly useful for those just starting on their analytics journey.

6)     Planning: Knowing the goals, the following step is to develop a strategy for reaching the goals. At the practical level, this involves figuring out what data that you need to obtain in order to answer the goals. Subsequently, how to obtain them, and then how to analyze them, reach actionable results, visualize these and report them to different types of stakeholders. This step has a number of sub-steps that each bear consideration, notably feature selection, message hierarchy planning, stakeholder requirements, analysis design and visualization. It may sound like a lot of time needs to be sunk into planning before one can even begin gathering data, but this is not the case: Planning complexity is directly proportional to goal complexity. If all you want to know is how many players that complete your game, planning is straightforward. If you want to be able to develop 3D heatmaps, player profiling through cluster analysis, or track 100s or 1000s of behaviors, you are looking at a more complicated knowledge discovery process which it will take time to plan properly. As mentioned above, adopting an evolutionary prototyping strategy – starting small then building up – is recommended for non-experts or when working with unfamiliar types of games (the kind of data that provide insights in Mario Cart, Farmville and Deus Ex are somewhat different).

7)     Collection: Having spent time formulating clear goals with the analytics process – what we want to achieve, while being open and prepared for being surprised! – and planned how to reach the goals, it is time to collect data. There are many ways to do this – building an in-house system, using third-party tools – and a lot of different ways to technically implement solutions. Here we are going to assume that some sort of collection system is in place, and provide some ideas for where to go look for more information on telemetry systems.

8)     Analysis: Having collected data it is time to analyse them. There are again many ways to handle this. Often simple statistical methods like calculating averages or sums will provide the needed information, in other cases multi-variate statistics or techniques from Machine Learning will be needed.

9)     Visualization: Having analyzed the data they must be visualized in a way that makes them understandable to the audience. If the audience is the analyst her/him-self, this can be relatively uncomplicated as they know the numbers, however, people in different professions tend to thin differently. For example, while an analyst will be happy with a data table, a designer might need a visual representation of the data in order to understand the result and the implications. A well-known example in the industry is the heatmap, which visualizes e.g. death events on top or inside of a virtual environment.

10) Reporting: Now it is time to take the results of all the hard work to the team, to the management, the colleagues in marketing, the user researchers, testers or whomever else the insights are for.

11) Implementation: With some cool results at hand, recommendations for how to change a game’s design, a work process or similar, can be formulated. This is rarely the job of a dedicated analyst – they report the issues, but leave the experts in the various domains to decide on solutions. For example, if it was found that players in a FPS had a hard time with a specific boss battle, the reason being they did not find a hidden rocket launcher, the analyst usually presents the result, but leaves it to the designers to figure out how best to make the rocket launcher more visible. Microsoft Studios Research are famous for building strong and successful connections between user research/user testing, analytics and design teams.

12) Maturity: Having gone through a first cycle of planning, collecting data, analyzing them, it is time to take a step forward – maybe to track more data, use more sophisticated analysis methods, experiment with interactive visualization of data, read up on analytics in other sectors of the industry and see if there are ideas useful to your specific situation, etc. The possibilities are in every sense of the word unlimited in scope.

In the next post, we will start with the first step: The basics →

—————————–