$2B+ of cargo moving through 40+ countries, and zero idea where any of it was, until now. A real-time visibility platform that turns a fog of carrier emails and EDI feeds into a single live map of every shipment.
Shipment status lived in carrier portals, EDI batches, and inboxes. Reconciling it took analysts hours, and by the time an exception surfaced the truck had already missed the dock.
We replaced that polling-and-spreadsheet workflow with an event-driven backbone: every carrier signal becomes a Kafka event the instant it lands, so the platform reflects reality in seconds rather than the next morning.
Thirty-plus carriers, each with its own format, cadence, and reliability, made request-response integration brittle. Kafka gave us replayable, ordered streams and let us add carriers without touching downstream code.
Redis materializes the current state of each shipment for instant reads, while PostgreSQL keeps the full ACID history. Reads stay fast under load, and nothing is ever lost.
| Component | Technology | Why |
|---|---|---|
| Event backbone | Kafka | Decouples 30+ noisy carrier feeds; replayable, ordered per lane. |
| Stream processing | Node.js consumers | Async I/O fits high-fan-in, low-CPU normalization work. |
| Live state | Redis | Sub-50ms reads for dashboards without hammering Postgres. |
| System of record | PostgreSQL | ACID history, JSONB events, and rich analytical queries. |
| Infrastructure | AWS (MSK, ECS, RDS) | Managed, multi-AZ, autoscaling across global regions. |
Every milestone from booking to delivery, stitched across carriers into one chronological view.
Geofence and telemetry signals continuously rescore arrival times as conditions change.
Delays, dwell, and detention trigger real-time alerts before they cascade into stockouts.
A canonical schema absorbs 30+ EDI and API formats so onboarding a new carrier is config, not code.
In-flight goods are reflected in stock positions, enabling leaner safety-stock policies.
Reconciled events push back into ERP and replenishment systems to close the planning loop.
“We didn't make the supply chain faster, we made it observable, and observability paid for itself in inventory.”
By turning a fragmented stream of carrier signals into one live, queryable picture of $2B+ in shipments across 40+ countries, the platform let planners trust their ETAs for the first time, and trust translated directly into a 23% inventory reduction. The Kafka-backed, Redis-materialized architecture sustains sub-50ms reads under global load while persisting a complete, auditable history in PostgreSQL, giving the business both real-time reflexes and long-term analytical depth on a single source of truth.