Case Study: Kolibri Games achieves 30% higher player lifetime value and rapid ML production with Databricks

Predicting and increasing customer lifetime value with ML

Kolibri Games, a Berlin-based mobile studio behind hits like Idle Miner Tycoon with over 10 million monthly active users, faced scaling and collaboration problems as data volumes and team size grew. The company needed to speed a two-week data release cycle, unify engineers and analysts, and build reliable machine-learning predictions of player lifetime value while keeping a player-led product approach.

By adopting Databricks on Azure—moving ETL/streaming into Delta Lake, standardizing on notebooks and Python, and using MLflow for model ops—Kolibri centralized and automated its data platform. Outcomes included a 30% increase in player lifetime value, threefold improvement in data-team productivity, a dramatic cut in ML production time (from two weeks to two hours), better targeting that lowered acquisition costs, and a scalable foundation for personalization, recommendations, and churn prediction.

Kolibri Games

Oliver Löffler

CTO and Co-founder