Corvus Logistics
Real-time visibility across 1,400 vehicles and nine depots
Position data is current to under 90 seconds. Empty-running fell 12% in the first full quarter after rollout.
- data latency, from 18 hours
- <90s
- reduction in empty running
- 12%
- depots on one model
- 9
Problem
Nine depots on four systems, with fleet position reconciled overnight in a spreadsheet and planners making decisions on data up to 18 hours old.
Solution
A streaming ingestion layer over the existing telematics and depot systems, feeding one operational data model and a planner-facing view.
Result
Position data is current to under 90 seconds. Empty-running fell 12% in the first full quarter after rollout.
Corvus did not have a data problem so much as a reconciliation problem. Every depot had accurate data about itself and no way to see anyone else.
How the work ran
- Event streaming from four source systems into a single operational model, with the sources left in place and unchanged.
- A conformed vehicle and consignment identity resolved at ingest, which was most of the actual difficulty.
- Planner tooling built with two planners in the room for the whole build, replacing the overnight spreadsheet.
- Depot-by-depot cutover over eleven weeks, with the spreadsheet kept running in parallel until each depot asked to switch it off.
What made the difference
The parallel run was the unglamorous part and the reason adoption held. Planners stopped using the spreadsheet because the new view was better, not because it was taken away.
“The thing that stood out was the handover. Our team had been running the platform for six weeks before the engagement formally ended, and nothing broke when it did.”
Daniel Okonkwo
VP Engineering, Corvus Logistics