Fleet telemetry mesh

A logistics operator had GPS, electronic logging, and warehouse scans in three clouds. Dispatch saw a map. Maintenance exported a CSV. Nobody could say, from one place, why a trailer was late — or whether the cause was a device, a dwell, or a missed scan.

The devices were staying. The vendors were staying. What they needed was a query path and an ownership model, not another dashboard that would be sunset when a contract ended.

We put a thin event mesh in front of the existing devices and kept the vendors. The query path dispatch actually uses — late trailers, last known scan, dwell versus movement — sits on data the operator owns. Models that predict dwell stay behind a measured baseline and a fallback to the raw event list.

Security review sat on the same tickets as the ingest work: which roles see which fleets, what is retained, what happens when a device key rotates. Maintenance stopped living in exports.

Late trailers have a cause in under a minute. Maintenance stopped exporting CSVs. The mesh is a product the operator owns. Dispatch and maintenance now argue from the same event, not from two screenshots.