Your Game, Your Data, Time To Play!

Due to the nature of the mobile gaming industry it is quite difficult to find users organically. Mobile game owners or publishers spend the majority of their marketing budgets on ads that are needed to acquire users until they arrive organically. In other words, the mobile game companies spend all their marketing budget for user acquisition. 

Monetization is the most important part the this job. So, how does it work?

At the beginning, targeting the right user segment and acquiring them is enough. Because, the mobile game company starts to make a profit after each player's spending exceeds (in-app purchase and rewarded video) the money spent on acquiring them. After installing the game to the users it is important to keep them in the game as long as possible (Higher Retention). So, the longer the user/gamers stay, the more profit they leave. 

On that purpose, we developed a churn/retention model to understand behavior of users for one of our customer. Mainly, they creates and publish casual games and their vocabulary puzzle game has 15M+ downloads. 2017 and 2018 on Google Play and the Apple App Store, it received awards in the category of most downloaded games in Turkey. The German version of the game was prepared in 2018 and started to be published in the German market.

Churn project of Vocabulary Puzzle Game captures the four different models;

Day-one churn, Day-7 churn, Day- 15 and Day-30 churn. 

As a result of the models, they realize that one of the user group has catastrophically high day-one churn rate (%90). Then this group of user coming from the one specific ad networks and location. After they recognize this, they directly  lower the acquisition bid for this ad network and start to acquire users much cheaper. 

On the other hand, they realized that they lost 71% of the users within 30 Days. So, for Day-30 churn, they targeted to find lower churn rate group (20%). On that purpose they developed a Decision Tree Model and discover the features of this user group. After this, they start to spend more money to acquire this kind of a users (Depending on ad network, gender and age). Because this kind of a users are valuable, they play longer and spend more money and watch more rewarded video than the others. After this network optimization  and user segmentation, their over all churn rate for 30-Day decrease 62% for this new user group. Which means 1250 people/users started to stay longer in the game.

In other case with more technically, 

After getting the player in the game, data begins to accumulate within the scope of the session, chapters, movements and events of the game. After that, they can predict when the player will stop playing (Predicted Churn Dates).

More technically, for this model we found that most important key points which are highly effective on churn behavior of players. 

Which are (From the most important to least);

Number of chapters they finish, 

Number of hints which are taken in every single session,

Number of events in every single session, 

Device platform (android or ios),

Number of session for days where users are active

Number of rewarded video for each session 

Average session count 

Number of wrong trial

In detail; 

For 7 Day Churn: Overall churn rate is 54% (54% of users stop playing game after 7 days. )

On the other hand, we found such a meaningful group that 94% of them stopped playing the game within 7 days. The features of this group are as follows

They play less than five chapters, for per session their event count is higher than 27.5, each session they use more than two hints and for each active they play at least three session. 

--

Chapter less than 5

Event per session >27.5

Hint per session  > 2

Session for per active days > 3

--

Depending on the this knowledge they can predict churn day of every single new user. They can calculate the probability of churn on the seventh day for each user after entering the game. In this way, they can generate different offers and campaigns for depending on the churn probabilities. The benefits of this result are: they start to give more hints to users who has higher probability to churn at seventh day. And also show less rewarded video to not annoy them. As a result of this activities, 7 Day Churn percentage for new user group  decrease 41% (Previously, it was 54%). Which means 2200 people/users started to stay longer in the game. Even they decrease the intensity of rewarded video, LTV (Live Time Value) of this group is 58% more than previous users.

Churn rate : %24

Hint per session  < 0.9

Chapters  > 6

Churn rate : %27

Hint per session > 0.9

Platform:  android

Avg.session time < 685

Chapters  > 11

Everyone is different in many ways; tastes, lifestyles, expectations…

Why the game content they play is not different?

Why the game's difficulty level or progression will be the same for everyone?

Starting with the motto "Why do we offer everyone the same game", we have developed different game content and difficulty levels for each player.  When two players start the game at the same time they can move forward with a completely different content.

On that purpose, we developed a Chapter Finishing Model to understand behavior of users for one of our customers. Mainly, they create and publish casual games and their vocabulary puzzle game has 15M+ downloads. 2017 and 2018 on Google Play and the Apple App Store, it received awards in the category of most downloaded games in Turkey. The German version of the game was prepared in 2018 and started to be published in the German market.

More technically, by comparing the success rates of the sections, we have identified the sections where users have difficulty in playing. That is, users were unable to play, so we were able to identify the times when they left the game (stop playing). Similarly, we were able to determine which sections were more enjoyable or which had higher income potential. This allows us to offer user-specific gaming experiences. In this way, faulty and software problematic sections were also detected and fixed.

Users are segmented according to their playing behavior (how they played the game and how long does it take to finish chapters) by using the machine learning algorithm. After that the segment of new users is determined by using the predictive modelling. So, the contents and difficulty level of each chapter were adjusted to keep them longer.

Thanks to these arrangements, they start to offer more enjoyable and individual sections to appeal to each user. Thus, the number of users who stop playing due to the difficulty of the episode decreased by 60%.

We know that you all dream of building the next iconic game that will be remembered by future generations, you generally spend a lot of time thinking about some of the fundamental business questions for a sustainable business.

How much can I spend in acquiring a new player? What is the potential value of this user versus another one? When will my players churn and what can I do to prevent it? And of course, building models that help to predict the lifetime value (LTV) of your game players.

We are here to take advantage of the benefits data can bring. Here, capability to detect and predict the behaviour of players as early as possible is the most important factor for promising mobile game.

LTV prediction and its period should based on the your game type. Here we applied 14-day LTV prediction for Hyper Casual Game while we apply longer time period for Casual Game such as 30-days, 60-days etc.

Predicting determinant attributes of users is valuable both in terms of reducing expenses and increasing revenue. These effect directly to the LTV prediction. An urgent need in F2P mobile games is to convert players from “Non-spending” to “Valuable” users to reach sustainable Life-Time Value. So firstly, AppNava detect “Valuable” users to predict LTV. Next step is predicting number of purchases. Last step is predicting the amount of purchase. All these three steps take you to successful LTV prediction.

The breadth of data generated by gameplay and the ability to target actions at a user level has led to customer analytics becoming a core part of a games business. 

We know that you all dream of building the next iconic game that will be remembered by future generations, you generally spend a lot of time thinking about some of the fundamental questions for a sustainable business.

How much can I spend in acquiring a new player? What is the potential value of this user versus another one? When will my players churn and what can I do to prevent it?

And of course, building the models that help to predict the monetization of your game players. We are here to take advantage of the benefits data can bring.

For hyper casual games, your monetization mostly depends on the revenue which is coming from the rewarded video. The most important issue to be considered here is that the frequency and amount of watched video is quite effective on your profitability.

As a result of the data analyzes, the inferences are as follows;

Frequent session entries and rewarded video watching are more profitable than a single session entry and multiple rewarded video watching. The marginal benefit obtained from rewarded video watching decreases considerably in consecutive vide watching.

Based on this finding, we define two types of user segment. One of them are labeled as “Valuable” who log in the game frequently and watch average 6 rewarded videos in each session. And of course we have “Not Valuable” who also log in the game frequently and watch average 2 rewarded videos in each session. Here we focused on the “Not Valuable”, because it would be much more profitable if they start to watch more rewarded videos. 

After that we have solved out how “Not Valuable” act in the game. How they play the game and how they will behave in the future. So we start to predict whether new users are “Not Valuable” or “Valuable”. If they are “Not Valuable”, the reward they receive after watching the video is improved.

As a result of this arrangement, average number of watched rewarded videos increased from 2 to 4. And of course the total profit from this group increased by 32%.

User acquisition and monetization is the most important part of the mobile game industry. After they acquire the users/gamers, they are targeting the obtain maximum profit from them. It is natural process, they pay to acquire them, now it is time to earn money and make profit.

So how will this happen?

If your game is freemium and want to survive, there are two ways to do this; rewarded videos and in-app purchase. But if you have a premium game, situation could be completely different.

Let's return our case and discuss the situation. As you know each player has different behavior in your game.

Some users prefer to move slower and watch fewer rewarded videos, while some of them are more aggressive. They prefer to watch rewarded video at pretty good rate and perform extra in-app purchases. On the other hand, there could be only rewarded video viewers or only shoppers (in-app purchase) among your users. There are also those who do nothing.

Now, imagine that you can determine what segment your users belong to when they first enter the game. Are they video viewer, shopper, or nothing?

Thanks to the power of this knowledge, one of our customers start to identify their users as they first enter the game. So, they started to offer special content to them, knowing which player to proceed. They have offered different prices to users who are more aggressive advancing for item and gem sales. For those who are calmer and only inclined to watch video, they propose in-app purchases with extra campaigns and discounts.

In this way, they increased their monthly profitability rates by 22% with the new campaigns and referrals implemented by different user segments.

Knowing the user and knowing what he/she wants is all you need. After that, the campaign, discount and the way you will earn extra profit is up to you.