The problem
Ship2Truck moves freight across multiple land borders. GPS routinely went dark for two to four hours per crossing - devices in metal containers, in customs yards, with patchy cellular and zero satellite line-of-sight. Customers staring at the dashboard during a critical hand-off saw a stale dot.
The visibility gap was the SLA-killing part of every shipment, and the part the dashboard was worst at explaining.
What we built
A hybrid tracking model that doesn't depend on continuous GPS. Position is inferred from carrier-provided checkpoint events, weighed against last-known signal, and presented honestly with a confidence band.
- Checkpoint ingestion from carrier APIs, customs systems, and partner handoff events.
- Confidence bands on the dashboard - the customer sees a range when we don't have a fix, not a wrong dot.
- Event-sourced state so the timeline can replay any container's day in under a second.
- SLA-aware alerting that escalates only when an inferred position threatens an actual customer commitment.
How we shipped it
Two weeks of architecture work to nail the event schema and confidence logic - that was the whole game. Build phase ran in two-week sprints with shadow traffic from week three. Cut-over was incremental, lane by lane, with clear roll-back at every step.
What we measured
- Visibility blackout minutes per shipment
- SLA attainment quarter over quarter
- Customer-initiated "where is my container" tickets
- Dashboard accuracy versus eventual ground truth
The result
Border-crossing visibility blackouts dropped 96%. Quarter-over-quarter SLA attainment moved 12 percentage points. Inbound "where is it" tickets fell roughly two-thirds. The engagement paid for itself inside two weeks of operation.
Operator's note
"They understood the operational reality before they wrote any code. The model they shipped feels like it was designed by someone who had stood in a customs yard." - Rajiv K., Head of Product, Ship2Truck