November 21, 2012 at 18:36Collection and analysis of statistics in free-to-play games
If you are developing free-to-play games, then you are probably interested in questions related to the collection and analysis of statistics. Why? Because statistics are an important component of the success of free-to-play games.
The purpose of my series of articles is to structure diverse information on this issue, pass it through the prism of our experience, and give recommendations on
The success of free-to-play games depends on how much players are involved in the game and are ready to buy in-game bonuses that take the game to a new level in terms of extra features and achievements. The longer a player is in the game, the more he thinks about playing offline, the more likely he is to invest real money in his promotion of the game. Of course, it is more difficult for developers to provide such a level of excitement in games than in a traditional pay-to-play model.
One of the secrets of the success of F2P games is that their design should be based not only on the creative component and “ingenious” ideas, but to a greater extent on the analysis of the behavior of players in the game, that is, on real data / statistics. At the same time, you can start (and need) an F2P game only with a part of the finished content, and manage the development of the game / modify the content based on the needs of the players and the popularity of certain features. This approach is called data-driven design, or "data-driven design." It is a cycle, at each iteration of which there are four stages depicted in the figure.
The acceptable percentage of content availability at the time the game is launched depends on the genre, concept, etc. But what certainly should be ready when starting any free-to-play game is a powerful and flexible system for collecting and analyzing statistics, as well as a system for testing various options for functionality / art / balance. Moreover, all the indicators that are planned to be analyzed should be clearly planned, and the tools for analyzing and visualizing data should be pre-selected, integrated and configured.
My article series will consist of three parts, which will address the following issues.
From my own experience I will say that at first when working with statistics I want to fix almost everything in the game: every click, every game result and every screen in the game is shown. The thesis may be as follows: the main thing is to collect everything and not to miss anything, but you can figure it out later. This approach does not work for several reasons.
You can save a lot on analysis if you collect only the statistics that are really important for making decisions on the future development of the game. To do this, you need to start planning for collecting statistics at the stage of developing the concept of the game. For example, for our games, we compile a table in which opposite each indicator it is written which hypothesis it tests and what improvement can be made on the basis of knowledge about it.
The statistics that are collected in games are conventionally divided into three parts:
The collection of statistics of the first type — business indicators — is best automated, since they are 90% the same for all F2P games. There is an impressive array of analytical services that provide convenient solutions with visual data visualization and simple integration. Most of these services are paid, but you can hardly do without them, since the invention of “bicycles” (independent implementation of collecting business indicators) carries risks, extra costs and a waste of time. Read more about analytical systems in the third part of the series of articles.
Perhaps the most difficult part is tracking the behavior of players, since this part is usually unique for each game and requires certain analysis tools (which will be discussed in the second part of the article series). There are no ready-made solutions that can be integrated into the game and immediately begin to receive the necessary statistics. There are companies that can outsource the collection and analysis of statistics (for example, GamesAnalytics Ltd). But we prefer to allocate resources for this in the development team itself.
Technical information is the statistics that are needed in order to make the game more stable and timely fix the technical problems of players.
This is an indicator of the "fascination" of the game, which indicates how many people play the game every day.
The DAU / MAU value characterizes the proportion of all players who play the game every day. The higher this value, the more players involved, the more likely that players will buy in-game content. It is believed that if the DAU / MAU is greater than 0.2, the game can be considered successful.
It is worth noting that this is an approximate estimate, since to accurately calculate the return of players, it is necessary to clearly separate new players from those who returned at predetermined time intervals (usually daily), take into account the traffic source and the promotions held. A cohort analysis, which will be discussed in the second part of the cycle, helps in a detailed study of these issues. This indicator of “fascination” is simple and gives a quick characterization of the game.
It is important to keep track of the% of "paying" players, as well as their demographic and other characteristics. Knowing their portrait, you can focus on this particular audience when developing new functionality in games.
I will give an example on one of our games. The figure below shows the percentage of people playing by age and the percentage of paying people among them. It can be seen that it is better to focus on middle-aged people (35 - 54), since it is they who tend to pay.
In addition, it is important to be able to distinguish “whales” among the players: these are the people who spend a lot of money. It is necessary to get to know these people closer, to study their characteristic patterns of behavior, in which place they fall off in order to satisfy their needs as much as possible.
Why "whales"? In general, sometimes all paying players are divided into minnows, dolphins and whales. Pescaras spend little - about $ 1 per month. "Dolphins" - about $ 5, and "whales" - a lot. According to Gigaom in Zynga games, the top 20% of “paying” players spend an average of $ 1,100 per year ($ 90 per month).
Income Indicators:
Virality is a way of distributing information about a game on the Internet and social networks from player to player. If the game has well-developed viral mechanisms, then the cost of attracting new users is reduced. To monitor virality, you can use the k-factor.
The k-factor can be calculated using the following formula: k = X * Y, where X is the number of invitations per player, Y is the percentage of people who accepted these invitations by joining the game. If the k-factor is 0.2, then for each new player you can get 0.2 players who came to the game by invitation (in other words: for every five new players, we get one free player who came into the game by invitation). It is clear that the higher the k-factor of the game, the cheaper it becomes to attract new players to the game.
The first thing you need to analyze the behavior of players is the statistics on the progress of players in the game. To track progress according to the game scenario, the control points that players must pass are determined. Analysis of the speed of advancement at these points, the parameters of the players at these points will help to identify obstacles or difficulties in the game that need to be eliminated.
If the player made the first purchase, then he is transferred to the category of "paying" players. It is believed that the first purchase is a psychological barrier, once overcoming which, players part with money much more easily. Plan ahead in the game the sequence of actions that may lead the player to the first purchase. Track how many players implement the scenarios you define, work on the conversion, improving the interface and balance.
If a player leaves the game during the tutorial, consider that this player is lost for you: with a high probability he will never return to the game. To avoid this, the beginning of the game should be maximally staged. It is necessary to track each step of the tutorial in order to understand on which screen the player got bored and left the game, it was not clear to him whether he was able to learn, whether he had completed the first task on his own.
Tracking the player’s first and last activity during a game session may be helpful.
The first event sets the tone for the entire gaming session. It can captivate the player and make you spend a lot of time in the game. But the first event may also “scare away” the player, as a result of which he will close the game and may not return. It is necessary to compare and test which events / windows / greetings lead to more time in the game.
The latter event is also important. The last event usually becomes exactly the obstacle in the game that should be removed. If the last event for the game session is planned (for example, the player is waiting for the completion of a certain game cycle), it is worth making this event so that the player would like to enter the game next time.
Since I’m developing mobile games, I’ll give an example, say, from an exciting Android world.
It can be useful to collect statistics on the technical equipment of player devices to ensure game stability. For example, it is important to know which devices, firmware, screen resolutions, types of hardware-supported textures are the most popular among players. It is also important to know which hardware configuration generates the most revenue and player return (the difference in revenue can vary by tens of percent). It is worth reducing the list of supported devices if they do not generate income and if the game is unstable on them. This, in addition, will protect the application from negative reviews in the store.
If the game uses resuming resources, collect statistics about successful resuming, about the number of resuming requests, about errors that occur during resuming. If the resume occurs before the first start of the game, then it can scare away a solid part of the audience. And if the players have not downloaded the game, then they certainly will not return and will not pay. Therefore, you need to take care of the maximum stability of the download procedure and find something for the players to wait for. And even better - to find the opportunity not to download data at the start, but to download inside the game for an additional fee.
If the game uses offer systems as an additional monetization, it makes sense to monitor the effectiveness of their work, including checking the coverage of offers in different countries on different devices.
A lot of useful information can be found in the documentation, presentations, articles prepared by the analytical services themselves. As a rule, he gives competent examples, cases, rationales, industry indicators. Here is a list of services that helped me deal with the issue of collecting and analyzing statistics in games:
Another useful article about developing free-to-play games:
The Design of Free-To-Play Games: Part 1
The Design of Free-to-Play Games, Part 2
Of course, the most basic metrics that are worth tracking when developing free-to-play games are listed here. But even they already provide a lot of information for deciding on the development of the game.
Do you have ideas for other important metrics? I will be glad to see your comment!
In the next article, I would like to dwell in detail on the main analysis methods that you need to own in order to extract really useful information from a sea of data. The main methods that will be considered: user segmentation, cohort analysis (the behavior of groups of people over time), funnels or analysis of transition sequences, A / B testing.
The purpose of my series of articles is to structure diverse information on this issue, pass it through the prism of our experience, and give recommendations on
- what indicators are worth tracking in games;
- What analysis tools can help with statistics?
- what statistics collection and analysis services exist with their advantages and disadvantages.
The success of free-to-play games depends on how much players are involved in the game and are ready to buy in-game bonuses that take the game to a new level in terms of extra features and achievements. The longer a player is in the game, the more he thinks about playing offline, the more likely he is to invest real money in his promotion of the game. Of course, it is more difficult for developers to provide such a level of excitement in games than in a traditional pay-to-play model.
One of the secrets of the success of F2P games is that their design should be based not only on the creative component and “ingenious” ideas, but to a greater extent on the analysis of the behavior of players in the game, that is, on real data / statistics. At the same time, you can start (and need) an F2P game only with a part of the finished content, and manage the development of the game / modify the content based on the needs of the players and the popularity of certain features. This approach is called data-driven design, or "data-driven design." It is a cycle, at each iteration of which there are four stages depicted in the figure.
The acceptable percentage of content availability at the time the game is launched depends on the genre, concept, etc. But what certainly should be ready when starting any free-to-play game is a powerful and flexible system for collecting and analyzing statistics, as well as a system for testing various options for functionality / art / balance. Moreover, all the indicators that are planned to be analyzed should be clearly planned, and the tools for analyzing and visualizing data should be pre-selected, integrated and configured.
My article series will consist of three parts, which will address the following issues.
- The main indicators that are worth watching in free-to-play games, and data on the behavior of players that should be analyzed to improve these indicators.
- The main methods of analyzing the collected data for making decisions on the development of the game: user segmentation, cohort analysis, “funnel” or analysis of transition sequences, A / B testing.
- Existing services with their advantages and disadvantages.
What statistics should be collected in F2P games
From my own experience I will say that at first when working with statistics I want to fix almost everything in the game: every click, every game result and every screen in the game is shown. The thesis may be as follows: the main thing is to collect everything and not to miss anything, but you can figure it out later. This approach does not work for several reasons.
- Analyzing huge amounts of data is expensive: you need to attract a lot of high-class analysts who must have advanced knowledge, both in statistics and in the methods of its processing, to be familiar with OLAP cubes, artificial intelligence algorithms, etc. That is, the less data - the better!
- The data quickly becomes outdated, since it depends on the marketing campaigns carried out, on the source of attracting players, on innovations in the game, and even on the time of year. Therefore, all indicators are important to watch in realtime.
You can save a lot on analysis if you collect only the statistics that are really important for making decisions on the future development of the game. To do this, you need to start planning for collecting statistics at the stage of developing the concept of the game. For example, for our games, we compile a table in which opposite each indicator it is written which hypothesis it tests and what improvement can be made on the basis of knowledge about it.
| Indicator | Decisions made |
| Income by levels and domestic products | If advanced players pay more, then you need to work to stimulate buying earlier (analyze needs at early levels, lower prices for some products, etc.). If you pay more at the beginning of the game, then you need to introduce special products for more advanced players, add an additional opportunity to spend the accumulated currency. |
| Points earned by players by level | The data will help set more adequate goals for the players, as well as adjust the game balance. |
| Game assignments | For each task there is an approximate estimate of how much time a player will need to complete it. By comparing the actual time to complete the task with the expected, you can adjust the parameters of the tasks and their sequence. |
The statistics that are collected in games are conventionally divided into three parts:
- business performance;
- player behavior;
- Technical information.
The collection of statistics of the first type — business indicators — is best automated, since they are 90% the same for all F2P games. There is an impressive array of analytical services that provide convenient solutions with visual data visualization and simple integration. Most of these services are paid, but you can hardly do without them, since the invention of “bicycles” (independent implementation of collecting business indicators) carries risks, extra costs and a waste of time. Read more about analytical systems in the third part of the series of articles.
Perhaps the most difficult part is tracking the behavior of players, since this part is usually unique for each game and requires certain analysis tools (which will be discussed in the second part of the article series). There are no ready-made solutions that can be integrated into the game and immediately begin to receive the necessary statistics. There are companies that can outsource the collection and analysis of statistics (for example, GamesAnalytics Ltd). But we prefer to allocate resources for this in the development team itself.
Technical information is the statistics that are needed in order to make the game more stable and timely fix the technical problems of players.
Business performance
DAU / MAU
This is an indicator of the "fascination" of the game, which indicates how many people play the game every day.
- DAU (daily active users) is the number of unique users who run the game at least once a day.
- MAU (monthly active users) is the number of unique users who run the game at least once a month.
The DAU / MAU value characterizes the proportion of all players who play the game every day. The higher this value, the more players involved, the more likely that players will buy in-game content. It is believed that if the DAU / MAU is greater than 0.2, the game can be considered successful.
It is worth noting that this is an approximate estimate, since to accurately calculate the return of players, it is necessary to clearly separate new players from those who returned at predetermined time intervals (usually daily), take into account the traffic source and the promotions held. A cohort analysis, which will be discussed in the second part of the cycle, helps in a detailed study of these issues. This indicator of “fascination” is simple and gives a quick characterization of the game.
Paying Players
It is important to keep track of the% of "paying" players, as well as their demographic and other characteristics. Knowing their portrait, you can focus on this particular audience when developing new functionality in games.
I will give an example on one of our games. The figure below shows the percentage of people playing by age and the percentage of paying people among them. It can be seen that it is better to focus on middle-aged people (35 - 54), since it is they who tend to pay.
In addition, it is important to be able to distinguish “whales” among the players: these are the people who spend a lot of money. It is necessary to get to know these people closer, to study their characteristic patterns of behavior, in which place they fall off in order to satisfy their needs as much as possible.
Why "whales"? In general, sometimes all paying players are divided into minnows, dolphins and whales. Pescaras spend little - about $ 1 per month. "Dolphins" - about $ 5, and "whales" - a lot. According to Gigaom in Zynga games, the top 20% of “paying” players spend an average of $ 1,100 per year ($ 90 per month).
Income Indicators:
- ARPU - average income per player (both paid and free installs are considered; the indicator is usually calculated per month).
- ARPPU - how much paying players spend on average (that is, the cost of the game in fact).
k-factor - viral coefficient
Virality is a way of distributing information about a game on the Internet and social networks from player to player. If the game has well-developed viral mechanisms, then the cost of attracting new users is reduced. To monitor virality, you can use the k-factor.
The k-factor can be calculated using the following formula: k = X * Y, where X is the number of invitations per player, Y is the percentage of people who accepted these invitations by joining the game. If the k-factor is 0.2, then for each new player you can get 0.2 players who came to the game by invitation (in other words: for every five new players, we get one free player who came into the game by invitation). It is clear that the higher the k-factor of the game, the cheaper it becomes to attract new players to the game.
Player Behavior Analysis
Player progress in the game
The first thing you need to analyze the behavior of players is the statistics on the progress of players in the game. To track progress according to the game scenario, the control points that players must pass are determined. Analysis of the speed of advancement at these points, the parameters of the players at these points will help to identify obstacles or difficulties in the game that need to be eliminated.
First Purchase Scenarios
If the player made the first purchase, then he is transferred to the category of "paying" players. It is believed that the first purchase is a psychological barrier, once overcoming which, players part with money much more easily. Plan ahead in the game the sequence of actions that may lead the player to the first purchase. Track how many players implement the scenarios you define, work on the conversion, improving the interface and balance.
Tutorial
If a player leaves the game during the tutorial, consider that this player is lost for you: with a high probability he will never return to the game. To avoid this, the beginning of the game should be maximally staged. It is necessary to track each step of the tutorial in order to understand on which screen the player got bored and left the game, it was not clear to him whether he was able to learn, whether he had completed the first task on his own.
Player’s first and last action
Tracking the player’s first and last activity during a game session may be helpful.
The first event sets the tone for the entire gaming session. It can captivate the player and make you spend a lot of time in the game. But the first event may also “scare away” the player, as a result of which he will close the game and may not return. It is necessary to compare and test which events / windows / greetings lead to more time in the game.
The latter event is also important. The last event usually becomes exactly the obstacle in the game that should be removed. If the last event for the game session is planned (for example, the player is waiting for the completion of a certain game cycle), it is worth making this event so that the player would like to enter the game next time.
Collection of technical statistics
Since I’m developing mobile games, I’ll give an example, say, from an exciting Android world.
It can be useful to collect statistics on the technical equipment of player devices to ensure game stability. For example, it is important to know which devices, firmware, screen resolutions, types of hardware-supported textures are the most popular among players. It is also important to know which hardware configuration generates the most revenue and player return (the difference in revenue can vary by tens of percent). It is worth reducing the list of supported devices if they do not generate income and if the game is unstable on them. This, in addition, will protect the application from negative reviews in the store.
If the game uses resuming resources, collect statistics about successful resuming, about the number of resuming requests, about errors that occur during resuming. If the resume occurs before the first start of the game, then it can scare away a solid part of the audience. And if the players have not downloaded the game, then they certainly will not return and will not pay. Therefore, you need to take care of the maximum stability of the download procedure and find something for the players to wait for. And even better - to find the opportunity not to download data at the start, but to download inside the game for an additional fee.
If the game uses offer systems as an additional monetization, it makes sense to monitor the effectiveness of their work, including checking the coverage of offers in different countries on different devices.
What to read on this topic
A lot of useful information can be found in the documentation, presentations, articles prepared by the analytical services themselves. As a rule, he gives competent examples, cases, rationales, industry indicators. Here is a list of services that helped me deal with the issue of collecting and analyzing statistics in games:
Another useful article about developing free-to-play games:
The Design of Free-To-Play Games: Part 1
The Design of Free-to-Play Games, Part 2
To be continued…
Of course, the most basic metrics that are worth tracking when developing free-to-play games are listed here. But even they already provide a lot of information for deciding on the development of the game.
Do you have ideas for other important metrics? I will be glad to see your comment!
In the next article, I would like to dwell in detail on the main analysis methods that you need to own in order to extract really useful information from a sea of data. The main methods that will be considered: user segmentation, cohort analysis (the behavior of groups of people over time), funnels or analysis of transition sequences, A / B testing.