FunCraft Inc. specializes in creating casual mobile games, such as word puzzles and logic games, which are primarily monetized through advertising. In the mobile gaming sector, the marketing department plays a key role as the company's financial growth engine, constantly striving to expand its user base.
The company is headquartered in the United States, with its team spread across the United States, Argentina, Spain, and Israel.
Effective management of marketing budgets is one of the key challenges in this dynamic industry. A particular difficulty is the long horizon of return on investment in marketing activities. The revenue generated from user acquisition spend depends on how long players remain active in the app. It often happens that the full return on investment occurs only after many months, and long-term user behavior differs dramatically between individual player acquisition channels, which means that only in the long run can marketers assess whether and which of their actions were profitable.
An additional challenge is the instability of the environment, in which the quality of users acquired by marketing partners changes, advertising rates fluctuate, and constant application updates affect both player engagement and monetization. In such conditions, budget management requires advanced tools that allow you to precisely forecast the profitability of marketing campaigns in the long term.
FunCraft needed a system to forecast the long-term ROI of marketing campaigns and assess the potential for results for new games and marketing channels with a minimal amount of data. It was crucial to speed up the decision-making process in order to avoid losses.
Scope of work
The project involved creating a flexible and accurate prediction system based on marketing data from Adjust and other sources that would respond to FunCraft's needs. Key stages:
- Data collection and quality assurance:
- Our team collected data from Adjust, Applovin MAX and store platforms (Apple Store, Play Store).
- We ensured the correctness of the data by eliminating deviations from the values in Adjust in the source of truth for data on advertising revenues, i.e. in the Applovin MAX mediation platform, and in the source of truth for revenues from in-app purchases, i.e. Apple Store and Play Store
- Building a ROAS predictive model:
- Model ML (XGBoost): We decomposed the LTV curve into a retention curve and ARPDAU, which allowed us to build a solution that was flexible and easily adaptable to changes in the environment.
- Analytical model: We have created an alternative statistical and econometric approach that is effective with very limited amounts of data, copes well with predictions for new games and marketing channels, is quick to implement and does not generate high maintenance costs.
- The models were created in Bigquery, allowing you to reduce the implementation cost by avoiding the need to create additional ETL processes
- Reporting in Looker Studio:
- As part of the project, our team created visualizations of ROAS predictions over a long-term horizon, featuring clear charts and analyses that highlight deviations from actual results, making model insights more accessible to non-technical users.
- We have developed a dashboard with investment recommendations based on forecast results, which makes it easier for marketing teams to identify the best campaigns and areas for optimization, reducing the time needed to draw conclusions.
- Analytical support:
- Our analysts interpreted model results and provided recommendations for budget decisions when the results seemed ambiguous to the marketing department in interpretation
- We explained the reasons behind the model results, providing the marketing department with detailed information that helped them understand the strengths and weaknesses of each marketing campaign.
Results
Key results:
- High model accuracy with minimal prediction error.
- Rapid implementation, delivering initial insights within weeks.
- Reduction of investment errors by identifying channels, markets and games with high and low ROI potential.
Business value:
- Greater ROI: Improved allocation of marketing budgets leads to a measurable increase in return on investment, resulting in significant annual revenue growth.
- Operational efficiency: Reducing the UA manager's working time thanks to the automation of data analysis processes.
- Long-term adaptation: Models that dynamically adapt to the changing market environment.
Customer review:
The solution from Alterdata allowed us not only to improve the effectiveness of our marketing campaigns, but also to assess the potential of new games faster and more accurately. Thanks to this, we could make more informed decisions.
The solution from Alterdata allowed us not only to improve the effectiveness of our marketing campaigns, but also to assess the potential of new games faster and more accurately. Thanks to this, we could make more informed decisions.
Summary
ROAS prediction project for FunCraft Inc. is an example of effective adaptation of analytical tools to the dynamic gaming industry. Our predictive model not only enabled the optimization of marketing budgets, but also became a key decision-making tool.
We invite other companies from the mobile apps sector to cooperate on similar solutions, ensuring precision, efficiency and long-term success.
