What actually counts as inspection evidence
By HoldField · 2026-06-10
Most quality disputes come down to a single question: can you show what you saw when you made the decision? A verdict — pass or fail — is a conclusion. Evidence is everything that would let a second person reach the same conclusion without being in the room. Those are not the same thing, and treating the verdict as the record is where traceability quietly breaks.
The four parts of a defensible record
- The observation: the image, measurement, or reading the decision was made from — not a description of it.
- The criteria: the specific rule or tolerance that was applied, versioned so you know which revision was in force.
- The decision: the verdict, plus who or what made it and when.
- The linkage: an unbroken chain tying the observation to the exact part, station, and moment.
If any of the four is missing, the record is an assertion rather than evidence. The most common gap is the first one: teams keep the verdict and discard the observation, so months later there is no way to review the call.
Why the observation has to be kept
Keeping the raw observation is what lets you re-examine a decision when a customer escalates, when a rule changes, or when a similar part fails downstream. It is also what lets you improve the criteria: you cannot tune a threshold you have no examples for. A record you can only read is documentation; a record you can re-examine is evidence.
A good test: pick a part inspected last quarter and try to reconstruct the decision from the record alone. If you can, it was evidence. If you are relying on someone remembering, it was not.
Sealing matters as much as capturing
Evidence that can be edited after the fact is weaker than evidence that cannot. This is why serious records are content-hashed at the moment of capture: any later change is detectable. You do not need a blockchain for this — a stored hash and a timestamp are enough to show a record is the same one that was made at inspection time.