Field drift shows up weeks before the warranty claim that reports it.

Signal

A cohort that separates, and the part that explains it.

Aggregate residuals by build month and most cohorts overlay each other almost exactly, which is what a healthy process looks like. When one separates, it separates cleanly, and it does so long before the population starts generating claims.

The cohort drawn here crosses the specification tolerance around week fourteen in service and is unambiguous by week eighteen. The corresponding warranty signal did not become statistically visible for another eleven weeks. That interval is the entire value proposition for a manufacturer: eleven weeks in which the cause can be found, the supplier engaged, and the affected population bounded before the claims curve turns.

Attribution to a component follows from the same data. Flag rate by part touched is unevenly distributed and usually dominated by one item — here a camera bracket at nearly nineteen percent against a fleet-wide baseline in the low single digits. That is a supplier conversation, not a vehicle conversation.

Scope

Campaign by measured drift, not by build range.

The conventional way to bound a campaign is by build window: every VIN produced between two dates gets recalled, because that is the only population the manufacturer can identify. It is safe, it is defensible, and it is enormously wasteful, because most of those vehicles are fine.

Field residuals allow the population to be narrowed by what actually happened rather than by when the vehicle was built. Of thirty-eight thousand vehicles in an affected cohort, six thousand show residuals above tolerance in service, and around four thousand of those are still in service and reachable. That is the set worth contacting.

The saving is real but the framing matters more than the arithmetic. A campaign scoped on measured field behaviour is a stronger position with a regulator than one scoped on a date range, because it demonstrates that the manufacturer knows which vehicles moved rather than which vehicles might have.

Feedback

Back into the tolerance, not just into the recall.

The durable value is not the campaign, it is what the field data says about the specification that allowed the problem. A tolerance that looked adequate on a test track and is not adequate after four years of thermal cycling and road input is a finding that belongs in the next drawing revision.

Signed field evidence makes that loop tractable, because the residuals arrive with their provenance attached — model version, thresholds, vehicle and cohort — and can therefore be aggregated across fleets and operators without laundering away the conditions they were measured under.

It is worth being explicit about our position in this. We do not design brackets, set tolerances, or validate builds. We produce a measurement of what happened in the field, in a form that survives being handed to someone who was not there.

Industries

The rest of the category.