Case Study: Bird detects offline scooters with Ververica and Apache Flink

Bird - Customer Case Study

Bird used Ververica’s Apache Flink-based stream processing capabilities to monitor scooter telemetry and detect when dockless scooters went offline. The challenge was that backfilling older Kafka data caused records to arrive out of order, making it difficult to reliably track scooter status and distinguish true offline events from false alerts.

Ververica helped Bird refine its Flink process-function design by moving watermark extraction into the Kafka source and then embedding buffering inside a keyed process function so event-time timers and incoming records were handled in a deterministic order. This eliminated false positives and missed events during backfills, produced consistent results for both offline detection and other jobs, and generalized into a reusable BufferedKeyedProcessFunction, though Bird noted the buffering can increase memory use and checkpoint size.

Bird

Philip Wilcox

Bird