An invoice proves a bay was paid. It does not prove the vehicle left correct.

Exposure

Billed is not verified.

Trace a book of collision claims through to outcome and the attrition is severe. Roughly two thirds touch ADAS components. Of those, about sixty percent have a recalibration billed. Of those, fewer than a third have any evidence that the calibration achieved anything.

The last step is where the exposure concentrates, and it is the one nobody currently measures because the invoice looks like proof. It is not. A recalibration invoice is a record that a bay was paid for a procedure — it says nothing about whether the vehicle's sensors were within tolerance when it drove away, and nothing at all about whether they still were a week later.

For a carrier the practical consequence is that a meaningful fraction of a book consists of vehicles that are documented as repaired and are not documented as correct. Those are different states, and only one of them is defensible in a coverage dispute.

Pricing

Latency is the variable you can actually move.

Expected loss from an undetected miscalibration is not linear in time. It rises fast in the first fortnight, while the vehicle accumulates exposure at highway speed with a sensor that is pointing slightly wrong, and then flattens once the risk has largely been met.

That shape matters for underwriting because it means the difference between detecting at three days and detecting at twenty-nine is most of the loss, while the difference between twenty-nine and sixty is comparatively little. Detection latency, not repair quality, is the lever with the steepest gradient.

Exposure is also not uniform across a book. It concentrates in vehicles old enough to have accumulated real drift and new enough to carry substantial ADAS content — a band that in most portfolios sits around six to eight years on L2-equipped vehicles, and which is growing every year as that content ages into it.

Recovery

The link nobody currently holds.

Subrogation against a repair facility requires establishing that the work was inadequate. The chain holds for the first three links — the incident, the claim file, the invoice showing calibration was billed — and then breaks, because no artefact exists that records whether the vehicle actually left within tolerance.

A signed evidence bundle is intended to supply exactly that link: an independent, timestamped, cryptographically verifiable record of the vehicle's sensor state after the repair, produced by a party with no interest in the invoice and verifiable years later without contacting us.

We are careful about how far to push this. Whether a given bundle supports recovery in a given jurisdiction is a legal question and not one we are qualified to answer. What we can say is that the artefact does not exist today, that its absence is the reason the chain breaks, and that carriers we have spoken to identify that specific link as the one they most want to close.

Industries

The rest of the category.