Logistics Company

The client operated across six vendor platforms — a warehouse management system, a transport management system, a customer portal, an ERP, a billing engine, and a third-party tracking API. None of them shared a data model. Reconciling shipment status across all six required two analysts spending two full days a week cross-referencing exports.

We mapped the canonical domain events the business actually cared about — shipment created, picked, dispatched, delivered, invoiced — and built an event-driven integration layer on Kafka. Each vendor system had a dedicated connector using Debezium for database-backed systems and lightweight polling adapters for API-only vendors.

The canonical event stream became the single source of truth. The customer portal read from it directly. The billing engine subscribed to delivery events instead of polling. Reconciliation reports that had taken two days to produce now ran on a cron job and took four minutes.

We instrumented the layer with OpenTelemetry from the start so the client could trace any event from ingestion to consumption. The operations team has been running it independently for over a year without incident.