The vehicle came back from the shop. Nobody checked whether it came back correct.

The gap

Measured in vehicle-days, not in sigma.

Drift is a statistics problem for us and a scheduling problem for you. What a maintenance lead actually feels is that a tractor unit sat in a bay for three days in April because the fault was found late, and that nobody could say in March that it was coming.

On a monitored fleet the same population loses roughly forty percent fewer vehicle-days over six months, and almost all of that comes from replacing unplanned bay time with a booked slot. Nothing about the calibration itself gets faster. What changes is that the work is known about a week earlier, which is the difference between a scheduled recheck and an asset stranded at a depot.

The baseline drawn on this chart is a counterfactual, not a measurement. We can show what a monitored fleet did; we cannot rerun the same six months without monitoring, and any vendor who claims otherwise is showing you a model with the labels filed off.

The arithmetic

Fifteen unnecessary rechecks are cheaper than one missed event.

Responding to a flag costs money whether or not the flag was right. Bay time for the recheck, the calibration if it turns out to be needed, the vehicle-day it consumes, and the administrative cost of scheduling it — call it thirteen hundred dollars in a typical US fleet.

An ADAS-related incident traceable to a missed recalibration is roughly fifteen times that, before anything is said about liability or a safety board. That ratio, and not detection accuracy, is what determines where the operating point should sit. It is why the scoring path is tuned for recall rather than precision, and why we would rather send you a recheck you did not need than miss one you did.

It also sets the honest break-even: if fewer than one flag in fifteen is real, monitoring is not worth what it costs you. Every pilot we run is scoped so that question gets answered with your own numbers rather than ours.

Operating it

Depot by depot, with a clock on every flag.

Nobody connects a fleet in a weekend. Rollout runs depot by depot, each one ramping over about two weeks as telematics identifiers are reconciled against repair-order records, and scoring starts at depot level as soon as that depot's data is clean rather than waiting for the last one.

Once it is running, the number that matters is not how many flags were raised but how many cleared inside the deadline. A hundred and sixty-eight hours, agreed in advance, with a named owner at every escalation rung. The weeks that miss are almost never detection failures — they are weeks where every bay was full, which is a capacity conversation rather than a software one.

That distinction is worth protecting. A monitoring layer that cannot tell you why the queue stopped moving is just another source of alerts, and a fleet that already ignores three dashboards does not need a fourth.

Industries

The rest of the category.