Introducing Windows in Kafka Streams - Kurly Tech Blog

Kurly’s SCM product team replaced batch-based settlement processing with an event-driven Kafka architecture.

They used Kafka Streams tumbling windows, grace periods, and suppression to aggregate inventory transactions by product code in 5-minute intervals.

The system now writes each final window result only once, reducing duplicate updates and database load.

The pattern helps turn bursty operational traffic into bounded, near-real-time stream aggregations for settlement processing.