Dropdown menu background

Highlights

Industry:

Mobile game development and publishing

Project scope:

  • Develop an AI-powered player lifetime value forecasting system
  • Combine advertising and in-app purchase revenue into a unified forecast
  • Automate revenue predictions across campaigns, markets, and platforms

The Client

A mobile game developer and publisher with an international audience and games available on mobile platforms. The company earns revenue through advertising and in-app purchases. Its teams use player analytics to guide marketing investment and product development.

The Project

The client wanted to understand the long-term value of newly acquired players earlier. This insight would help its teams compare acquisition channels, assess promising audiences, and plan advertising budgets. XIM developed a mobile game LTV prediction system that analyzes player activity and estimates future earnings. LTV, or lifetime value, represents the revenue a player generates over their time with a game. The solution brings engagement, retention, advertising revenue, and purchase behavior together. As a result, teams can connect early player activity with longer-term revenue potential. The system supports forecasts across several time horizons. It also updates those forecasts as more data becomes available, keeping revenue expectations aligned with observed player behavior.

Project duration:

8 months

Project team:

A small team, including a business analyst and a system developer

Project labor costs:

8 person-months

Technology stack:

Python, Pandas, Scipy, Statsmodels

The Solution

Mobile game LTV prediction from player behavior

XIM developed complementary machine learning models that analyze player engagement, retention, purchase behavior, and advertising activity. Each model estimates a specific part of future player value. The system then combines their outputs into an overall revenue forecast.

Individual predictions are grouped into cohorts—players with shared acquisition characteristics. This allows teams to compare expected value across campaigns, traffic sources, countries, and platforms.

Advertising and purchase revenue forecasting

The solution forecasts advertising revenue and in-app purchase revenue separately, then combines them into total lifetime value. This gives teams a clear view of how different audiences generate revenue and helps them assess acquisition opportunities.

As more player data becomes available, the system updates its forecasts. Once a forecast period is complete, actual revenue replaces the estimate for that period. Analysts can therefore review observed performance alongside expected future earnings.

Automated reporting and model validation

XIM built an automated service that retrieves fresh data, calculates forecasts, and stores the results in ClickHouse for analytics dashboards. It also preserves forecast history, allowing analysts to compare earlier predictions with actual revenue.

The team trained the models on historical data and evaluated them on separate datasets. This checked their performance on previously unseen cohorts and supported reliable use in the client’s analytics workflow.

The Outcome

XIM delivered an operational mobile game LTV prediction system that gives the client an earlier view of player revenue potential. Evaluation showed strong forecasting performance across most tested cohorts, with consistent results across repeated model retraining. Automated updates also made the forecasts part of the client’s regular analytics workflow. User acquisition teams gained a clearer basis for comparing traffic sources and planning budgets. Meanwhile, monetization and product teams gained insight into how player engagement translates into expected revenue. The solution brings advertising and purchase forecasts together in a practical tool for growth planning. It helps game teams evaluate acquisition opportunities with a clearer understanding of long-term player value.

Success stories

Hire our experts

Please enable JavaScript in your browser to complete this form.
Checkboxes