How Much Revenue Are We Making Right Now?

Immediately after a game launch, quickly grasping user access status and revenue performance is critically important. However, the existing environment based on Databricks and Elastic Stack had limitations, such as difficulty checking detailed metrics by country or OS, as well as data aggregation delays and high operating costs.

To solve this, Devsisters newly designed a near real-time metrics pipeline. This pipeline meets the following key requirements:

• Guaranteed latency under 10 minutes (SLO)
• Providing core metrics across various dimensions such as country, OS, and store
• Actively collecting client data using heartbeat logs
• Implementing complex business logic (such as reflecting exchange rates) through SQL-based ETL

The new system consists of streaming data ingestion via Kafka, followed by preprocessing, aggregation, and visualization stages. Through this, an environment was established where not only data engineers but also data analysts can use SQL to serve accurate and rapid metrics.