What are professionals talking about? Based on a round table on analyzing gaming projects as part of White Nights Moscow

In October, a team of mobile and web application analytics service devtodev visited White Nights 2016 - a cool international conference in Russia with the participation of world leaders in the gaming industry. The conference was held in Moscow and brought together more than 3000 industry professionals, including such giants as Google, Facebook, Unity, Amazon, Rovio and VKontakte.

I was fortunate to take part in a round table on game analytics with experts from Pixonic, Game Insight, OrcWork, Nevosoft and AppLovin - this event was the culmination of the first day of the conference.

For those interested, we collected answers to some questions addressed to all participants of the round table, and issued them in the Q & A format. Special thanks to Ruslan Gerasim and Yulia Lebedeva (Nevosoft) for help in preparing the material.
')
Q: Is it necessary for a person who is involved in gaming analytics to know mathematics, to be able to take an integral, to solve examples?

A: A company should have a dedicated analyst, or at least a person with the role of an analyst. He should have knowledge in the field of statistical mathematics at least. But more importantly, he must have deep analytical skills. Because the study of any analytics is very similar to the work of a detective. Quite often there are situations when it is necessary to evaluate according to the abundant, but few data why users act here in this way, and not as the game designer expects.

That is, we need not mathematics, but the ability to use it. You can also compare the work of the analyst with the work of Dr. House. When you have a bunch of disparate symptoms, and you need to make a diagnosis of "what is wrong in your game, what steps need to be taken to change it."

All people who are at least somehow connected with analytics, who make decisions on the basis of analytics, should understand at least the concepts. Those. they may not know how to build a specific mathematical model, how to take the logarithm, for example, but an understanding of the foundations of statistics and analytics is necessary for making a decision.

Q: What mobile analytics systems do you use? Own or others? One or more?

A: Participants mentioned systems such as devtodev, Amplitude, AppsFlyer. In Pixonic it happened historically: for six years they have been writing their analytics system, but Philip Gladkov recommended new companies to use ready-made solutions. In general, the use of several analytics systems at the same time is an absolutely correct approach, giving more valid data.

Q: Is it possible to determine by the user's behavior his gender, age without collecting his personal data?

A: You can use the Insights service from Facebook, see the audience of any game, any interests, and find out what kind of cars this audience drives.

You can also watch competitors, analyze their groups in social networks, create a player's portrait. Yes, and for your game, you can also try to create a portrait of a person who will play what you are doing. By the way, this is a good case when you look at social network players who are playing your game. Sometimes a portrait appears very well. After going through at least fifty portraits, you will form a fairly clear idea of ​​who this player is.

Q: Are there any other valid and important KPIs for f2p games other than LTV and advertising prices?

A: LTV and advertising price is the tip of the iceberg. Do they determine whether it is Titanic, crashes, crashes, goes, does not go? If LTV> CPI, then everything is fine.

If you continue to decompose the metrics, detail the question, then there will be retention, ARPU, ARPPU, and so on.

Here is the question - why do we need metrics? If you publish a game, you need to know LTV, CPI - that's enough. If we evaluate deeper - we take retention, ARPU. If you need to analyze a particular feature specifically, deeper metrics are used there, up to “how many battles are done in one session”.

Q: Is the behavior of users in different countries very different? Can the experience of one country be transferred to another in terms of large-scale game changes?

A: Differ greatly. For example, users in Japan can look twice at the game and already make an internal purchase, but in Russia this is unlikely. Therefore, using the data of two completely different countries and taking some results from them is very dangerous.

Another interesting case is related to the launch of the game Mushroom Wars in Korea. For us, this game seems difficult, and the Korean players are very fond of increased complexity. They flew through the levels on “hurray!”, And even those levels with which the developer himself is experiencing difficulties were too easy for Koreans. And I had to manually tighten the complexity of the game for them.

And, for example, when Game Insight launched the game Guns of Boom in Turkey, they noticed that the online subsided heavily in early September when schoolchildren ended the holidays. There were also noticeable drawdowns during the namaz.

During soft launch, it is better to focus on those countries that look like you are interested in: for the United States this is Canada, for Japan - Korea and so on. But to interfere with all the countries in one pile is impossible.

A video of the round table can be seen here .

From myself, I note that communication with industry experts contributes to the exchange of knowledge and experience, but the conferences are still not very often, and by participating in them, you still do not get structured knowledge, so we created an open course. ”Marketing and product analytics of games ” And invited experts from Game Insight, Pixonic, AlternativaPlatform (Tanks Online) and Nevosoft. If you are interested in analytics of game projects, and strive to improve your skills, join