in Blog

June 03, 2019

Mobile Gaming Analytics: Increase IAP, LTV and Retention using AI [Case Study]


Artur Haponik

CEO & Co-Founder

Reading time:

9 minutes

Popular mobile games can attract millions of players/users and generate terabytes of game-related data in a very short period of time. Such companies require scalable data analytics services and infrastructure to provide timely, actionable insights in a cost-effective way and with a very quick time response. Above all, for mobile gaming analytics infrastructure should be powerful, scalable and be able to process huge amounts of data and train Machine Learning models on top of that.

In order to meet all those above-mentioned needs, a growing number of companies use cloud infrastructures (Google Cloud or AWS) or internally build powerful data clusters. For instance, companies use advanced Web-scale analytics services to create a personalized experience for their players. Gaming companies use in-game telemetry data and smart Machine Learning instruments and algorithms to gain insights on how players engage and interact with the game. Those insights are very helpful to answer following questions:

  • At which game level do players get stuck?
  • What virtual products or goods did they buy?
  • What kind of actions pushed them to purchase?
  • Which gamers are likely to stop playing?
  • What is the best way to tailor the game to appeal to both casual and hardcore players?
  • Which players will spend money in the future?
  • How interaction within the game and and payments are correlated?

Mobile game publisher business challenges

Mobile game applications generate huge amount of game-event and player-telemetry data. This data has potential to provide insights into player behaviour and their engagement within the game. The nature of mobile games is a large number of client devices, a lot of in-game interaction, purchases and transactions. It means that mobile game analytics face unique challenges.

You may also find it interesting – Big Data in Mobile Gaming.

Customer profile and challenges

Our customer is a mobile gaming publisher and developer who issues a lot of different mobile games and acquires users all around the world. More than 500 people on board, 120k new users from different channels (Facebook, Google, Organic etc.). Those users generate terabytes of customer data from all over the world.

Optimized advertisement and customer acquisition budgets

Customer advertisement expenses (user acquisition budgets) were strongly correlated with the number of newly acquired customers. The rule was very simple: the more users they want to acquire and retain – more money they should spend on advertising. Our client data analytics and performance marketing departments struggled with marketing expense optimization, data processing and cost-effective data storage.

The cost of acquiring new customers was tremendous and only a small group of customers made purchases inside the app. That is why it is very important to quickly identify users, who will make a purchase in the future.

Customer retention

In short, within 30 days after app installation, 90% of users stopped using the app. A lot of money was spent on user acquisition but a high number of paying users were lost. It means that solving customer retention problems will definitely help the company to optimize marketing expenses and retain more paying users.

Unique user preferences and experience

That is to say, each user has preferences. Also each person plays differently: some of them like to quickly move to the next round and some play calmly. That is why important to have algorithm, which can predict the behaviour of each individual player.

Mobile gaming analytics challenges

After our customer’s application started gaining more and more players, the number of users that will upgrade to a paid account and start spending money or will refer app to friends increased. Additionally, there are other KPI’s that can give a possibility to understand gamers` behaviour and how they interact during the game. Games can and should be improved and updated using extracted insight from users` behaviour hidden in collected big data sets. As a result, user experience can be improved and gameplay time could be expanded using insights from Game Analytics solutions.

KPI’s to measure

To properly measure success of a game, our customer needs to understand how many users are active monthly? How many users you acquired during the last month? Or even how many users are paying within the game? Basically, there was a need for KPIs measuring. Examples of such KPIs are:

  • MAU (Monthly Acquired Users),
  • DAU (Daily Active Users),
  • ARPU (Average Revenue Per User)
  • IAP (In-App Purchases)
  • CAC (Customer Acquisition Costs)
  • MR (Monthly Revenue)
  • YoY (Year over Year)
  • and many other

Above measures can help to understand the performance and success reasons of your game and acquired users quality.

Solution – Mobile gaming analytics, AI And Big Data… Equals big revenue

Transform Big Data into an appropriate format

After a workshop session together with the business and technical team, we established a high-level roadmap of how to leverage their data and what kind of solutions could be applied. Addepto data engineering team have built a Data Lake with Data Warehouse to collect all the available data and transformed it into an appropriate format for data analysis. Collected data contains information on customers and their activity.

We decided to collect data on all customers’ in-game activity. We now track such data points as – progress, likes, interactions, preferences, expenses, demographics, behavior patterns and many other data points.

The mobile gaming analytics goal is a data monetization

Our solution was architected to find the most important factors, dependencies, correlations between the huge amounts of transactional data, and let game publishers monetize this data and improve gamers’ experience.

Machine Learning in mobile gaming analytics

Consequently, our solution contains Machine Learning models to understand users within terabytes of data. Similarly, Machine Learning models automatically perform an intelligent analysis of each user. It automatically extracts patterns in the data and applies them to millions of users.

ML algorithms speed up the data monetization process by quick analysis of millions of customers and find rules that can be used for prediction. Also, these algorithms cyclically learn on new data and adapt to current behavior patterns of customers.

Fraud detection in player acquisition

Our team of Data Scientists created Fraud Detection models, which identify bots using machine learning. For example, it reduces customer acquisition costs and focuses only on good quality customers. Furthermore, models used data from Tenjin combined with gaming server.

Solution produces an automated report on fraudulent bot activities and it’s streams. Report us delivered straight to the email box.

Customer lifetime value prediction (CLTV or LTV)

Our team has built and deployed into production LTV prediction AI model, which was integrated with marketing and advertising systems.

We implemented LTV machine learning models to automatically predict how much money particular customer will spend in your application. This information is used in segmenting new users in combination with customer acquisition cost (CAC). Our customer now able to choose only those clients that will generate relevant revenue in the future. That solution saves money on advertising among customers who will never buy anything.


Customer churn prediction

Customer Churn prediction machine learning models were implemented to predict, which paying customers are risky to stop using the application. In other words, receiving such information in a timely manner allows our customers to take action to retain more users.

In short, the highest value to the client was achieved when we connected churn prediction system with the marketing automation software and CRM systems. Now it sends personalized messages with tailored made offers per each customer automatically.

customer churn prediction dashboard

Recommendation engine using machine learning

For example, Addepto team deployed a Recommendation system to increase IAP (in-app purchases). For each of the players system personalizes offer and content, it is another similar game that user will probably want to play or a proposal for additional purchase to reach its goal in the game faster. That method increased customer satisfaction from the game, which directly contributed higher. business revenues.

Mobile gaming analytics – modern data warehouse and BI implementation

The integration of Machine learning results with original data into Data Warehouse gave a unique possibility for multidimensional data analysis. Also, very useful was a flexible self-service Business Intelligence (BI) solution. Using BI analytics, team was able to easily make any dashboard in a few minutes, set up alerts on preferred events or be emailed when something unusual is happening. Programming knowledge here is not necessary, those tools are fully self-service. Insights were easily shared with colleagues inside the company.

Marketing Performance and use Acquisition department are able to analyze risky users, their future revenues and create automated pipelines for targeting the right customers.


Results and impact of mobile gaming analytics solutions

Big Data playing an increasingly larger role in how the gaming industry collects and analyzes data. This intersection between gaming and analytics has resulted in the ability of gaming companies, such as Electronic Arts, to increase advertising revenue, improve gameplay, and efficiently manage the user experience.

Addepto’s customers Analytics team has now enabled advanced reports & dashboards to predict customers attiring from the business. That is to say, by unlocking insights, the management team initiated the right business decisions to prevent customer churn and personalize ad offers and increase game monetization.

Above all, key business decisions are made now based on insights from predictive models and analytics system

  • Develop a sustainable and robust strategy for customer retention and acquisition
  • Formulate plans to reacquire customers who have been left
  • Convert low-revenue earning customers into highly profitable ones
  • Reduce customer defections and improve profits
  • Track customer satisfaction by product, segment, and cost to serve
  • Increase IAP purchases and user experience

Each above-mentioned activity and interaction requires AI consulting. That is why it is worth monitoring how retention and LTV changes over time, what group of users are the riskiest and how users monetization goes. Likewise, it is also important to get alerts on such events.

AI consulting services

In conclusion, if you have any questions regarding the above project, implementation or results just ping us a message and we will be glad to tell you more about benefits and deployment.

mobile gaming analytics



Artificial Intelligence