Transport
Transport & fleet
Any model over a fleet is only as good as the position data underneath it. Where the integration did not exist, we wrote the protocol stack ourselves.

The work as it is done today
The operational questions in a transport business are simple to state and hard to answer: did it arrive, is this trip billable, who was closest, why was this run late. Every one of them is a question about position over time, and every one of them is usually answered by asking a driver.
Why we wrote the protocol stack
The devices in the vehicles spoke a standard we needed to consume, and the integration we needed did not exist in a form we could rely on. So we wrote it: TCP ingest, codec, storage and a live map, from the protocol up.
This is not something to do for its own sake. It was the right call here because the alternative was building an analytics layer on top of data we could not vouch for. A model over a fleet inherits every gap in the position feed underneath it, and those gaps are invisible in the output — the answer still looks like an answer.
What this makes possible
Once position is measured rather than reported, the operational questions have answers that hold up in a dispute. Arrival is a fact rather than a claim. A billable trip is derived from the same record the customer can be shown. And anything built on top of it — routing, exception detection, driver-facing tooling — starts from data that does not need to be apologised for.
What we would say before starting
If a device vendor already offers a reliable, documented feed, use it. Writing a protocol stack is justified when the integration is the constraint, not when it is merely inconvenient. We would tell you which case you are in before anyone signs anything.
Next industry
Retail & member operationsIs this your industry?
If the process above sounds like yours, describe what happens today. We will tell you which part a model can carry.
Send a request