Prediction in the gaming industry. Part 2: How well do you know your players
- Transfer
After you have a little understanding of the basics, we will consider in more detail what the benefits of forecasting for game developers are. Forecasting can become a kind of crystal ball for you, but only if you have the right information. So the first thing you need to do is start collecting data, a lot of data. If you do not have the ability to download data from the game, there will be no business.
This material is the second part of a series of articles that discuss the forecasting and use of predictive metrics in the gaming industry. If you have not read the previous article on forecasting, it is best to first read it here to get an idea of predictive analytics, modeling and reliability assessment.
As soon as the relevant data is at your disposal, it is the turn of analytics. There are a lot of techniques for thorough data analysis, for example, analytics based on indicators such as the daily number of active users (DAU), average user income (ARPU), and average session duration. Customer Value Index (LTV) is the most suitable for analyzing game data. In general terms, it represents the amount that the user is likely to spend before abandoning the game. Some consider the current expenses of the user, but we are more interested in either general or future expenses.
The most important thing is to understand which players are most valuable to you. They are all different: for every hardcore player whose LTV is measured in thousands, there are much more users who haven’t invested a penny in the game. And it’s in your interests as a developer to strive to retain players with a high LTV score.
But back to the metrics. LTV is great for getting a general idea of the situation (unless, of course, you have a reliable formula). But at the same time, this is a rather superficial way to evaluate your audience and its interaction with the game.
If you are developing games, then most likely you are now shaking your head disapprovingly. After all, users do not play in a confined space! They interact and communicate with other players, they have clans, guilds, social networks. Does that mean anything?
Such data are extremely important for forecasting and are known as social value. Like LTV, social value is also measured in dollars. Try to imagine that inside, for example, Facebook, there are thousands of social networks that users create around them. These are their friends and friends of friends. The traditional LTV metric involves an analysis of each network node to determine which players are most important to you, while the social value is determined by the network as a whole and the interaction within it.
Suppose you have a player (let's call him “player A”) who spends $ 5 in a game and player B who spends a dollar. Whose value is higher? A traditional metric would point to player A. But the analysis reveals that for every dollar spent by player B, player C receives $ 3, another $ 2 from player D, and another 3 from player E.
Add it all - and you look at player B in a completely different way.
Forecasting is a scientific technique, but do not underestimate the importance of human influence. Social value as a parameter takes this factor into account. Add up the LTV and social values and you’ll get the full value of the player. This option combines the best of two approaches. And he will help to understand which players are the most important for you.
It is worth noting that the most valuable players are often not the ones you think about in the first place. Users who spend a lot of money in the game and who have a high LTV score usually have no influence. And players whose net worth is determined by social connections and LTV spend a little, but have a significant impact on other players and thus trigger a chain reaction. These are the so-called “whales” (Social Whales), and it is precisely on them that one needs to be guided as a target audience.
Therefore, for game developers, the question for a million dollars (or even more, depending on the game) sounds like this: which player will quit playing, and how to prevent this? As you probably guessed, predictive analytics will help answer this question.
The predictive model not only reveals who has already left the game, but also shows who is on the verge of leaving and what kind of losses it faces. Of course, you cannot stop all users. Here we will come in handy such a parameter as the coefficient of outflow of users. The calculation of the expected amount of loss is as follows: the total cost of the player is multiplied by the degree of probability of his departure from the game. For example, the total cost of a user named Bob is $ 100, and we are 65% sure that he will leave the game. If we had a thousand such Beans, we could argue that over time the amount of loss from leaving any of them will average $ 65. Thus, we ascribe this amount to our Bob and, on the whole, are right.
Based on these data, conclusions can already be drawn. If the risk that an unremarkable user leaves the game is equated to the average, he can be let go. But what if the whale is on the verge of leaving? In this case, you need to seriously reconsider the strategy of retaining players, because with the departure of the “whale” you will lose its entire social network. Think of the expected amount of loss that you will incur if this happens: in the case of the same Bob, it will be $ 65.
The opposite of the outflow of users is conversion - a phenomenon when your players start investing in the game and the project makes a profit. Just a dream of any developer of free games. Basic analytic programs can identify users who have started making in-game purchases, but predictive analytics can do more, namely predict which players will start to do this and how much revenue it will bring to you. If you develop the system correctly, you can also predict which players will become “whales” and whom you should focus on in order to attract other users.
All of the above is by no means a universal guide to predictive analytics, even within the gaming industry. Over time, forecasting will reach a level where developers can predict the behavior of players in terms of virality, monetization, and even reimbursement of advertising costs.
So what awaits the gaming industry? I will talk about this in the final article of this series.
This material is the second part of a series of articles that discuss the forecasting and use of predictive metrics in the gaming industry. If you have not read the previous article on forecasting, it is best to first read it here to get an idea of predictive analytics, modeling and reliability assessment.
As soon as the relevant data is at your disposal, it is the turn of analytics. There are a lot of techniques for thorough data analysis, for example, analytics based on indicators such as the daily number of active users (DAU), average user income (ARPU), and average session duration. Customer Value Index (LTV) is the most suitable for analyzing game data. In general terms, it represents the amount that the user is likely to spend before abandoning the game. Some consider the current expenses of the user, but we are more interested in either general or future expenses.
The most important thing is to understand which players are most valuable to you. They are all different: for every hardcore player whose LTV is measured in thousands, there are much more users who haven’t invested a penny in the game. And it’s in your interests as a developer to strive to retain players with a high LTV score.
But back to the metrics. LTV is great for getting a general idea of the situation (unless, of course, you have a reliable formula). But at the same time, this is a rather superficial way to evaluate your audience and its interaction with the game.
If you are developing games, then most likely you are now shaking your head disapprovingly. After all, users do not play in a confined space! They interact and communicate with other players, they have clans, guilds, social networks. Does that mean anything?
Such data are extremely important for forecasting and are known as social value. Like LTV, social value is also measured in dollars. Try to imagine that inside, for example, Facebook, there are thousands of social networks that users create around them. These are their friends and friends of friends. The traditional LTV metric involves an analysis of each network node to determine which players are most important to you, while the social value is determined by the network as a whole and the interaction within it.
Suppose you have a player (let's call him “player A”) who spends $ 5 in a game and player B who spends a dollar. Whose value is higher? A traditional metric would point to player A. But the analysis reveals that for every dollar spent by player B, player C receives $ 3, another $ 2 from player D, and another 3 from player E.
Add it all - and you look at player B in a completely different way.
Forecasting is a scientific technique, but do not underestimate the importance of human influence. Social value as a parameter takes this factor into account. Add up the LTV and social values and you’ll get the full value of the player. This option combines the best of two approaches. And he will help to understand which players are the most important for you.
It is worth noting that the most valuable players are often not the ones you think about in the first place. Users who spend a lot of money in the game and who have a high LTV score usually have no influence. And players whose net worth is determined by social connections and LTV spend a little, but have a significant impact on other players and thus trigger a chain reaction. These are the so-called “whales” (Social Whales), and it is precisely on them that one needs to be guided as a target audience.
Therefore, for game developers, the question for a million dollars (or even more, depending on the game) sounds like this: which player will quit playing, and how to prevent this? As you probably guessed, predictive analytics will help answer this question.
The predictive model not only reveals who has already left the game, but also shows who is on the verge of leaving and what kind of losses it faces. Of course, you cannot stop all users. Here we will come in handy such a parameter as the coefficient of outflow of users. The calculation of the expected amount of loss is as follows: the total cost of the player is multiplied by the degree of probability of his departure from the game. For example, the total cost of a user named Bob is $ 100, and we are 65% sure that he will leave the game. If we had a thousand such Beans, we could argue that over time the amount of loss from leaving any of them will average $ 65. Thus, we ascribe this amount to our Bob and, on the whole, are right.
Based on these data, conclusions can already be drawn. If the risk that an unremarkable user leaves the game is equated to the average, he can be let go. But what if the whale is on the verge of leaving? In this case, you need to seriously reconsider the strategy of retaining players, because with the departure of the “whale” you will lose its entire social network. Think of the expected amount of loss that you will incur if this happens: in the case of the same Bob, it will be $ 65.
The opposite of the outflow of users is conversion - a phenomenon when your players start investing in the game and the project makes a profit. Just a dream of any developer of free games. Basic analytic programs can identify users who have started making in-game purchases, but predictive analytics can do more, namely predict which players will start to do this and how much revenue it will bring to you. If you develop the system correctly, you can also predict which players will become “whales” and whom you should focus on in order to attract other users.
All of the above is by no means a universal guide to predictive analytics, even within the gaming industry. Over time, forecasting will reach a level where developers can predict the behavior of players in terms of virality, monetization, and even reimbursement of advertising costs.
So what awaits the gaming industry? I will talk about this in the final article of this series.