We still have them. The labels came from looking at images together, not from an agreed physical inspection.

I've just asked for physical judgement on our trim plates for the same reason. No result to report yet, but I would keep the original disputed examples identifiable.

Can you link each example to its particular bezel, or only to the tray it came from?

Individual links exist for the selected examples. The larger image folder only has tray names, so I will keep those out of the reference decision.

Jamie, yes, but don't infer which from the moving streak alone. Record what actually changed while keeping the surface classification separate

They removed and replaced the bezel. Holder position wasn't recorded. That limits what I can recover from the existing images.

Could the next capture sheet name the seating and holder setup? Doesn't repair the old record, but it makes the next comparison interpretable.

And the lighting and exposure settings. If those change during reseating, a quieter streak cannot be assigned to the holder alone

I'd include the person who normally reseats them in the trial planning. An inspection arrangement that only its designer can reproduce will be awkward to hand over.

Keep the first set small enough to inspect properly. Hundreds of additional images won't compensate for not knowing which physical mark you're trying to reveal

Quality has supplied decisions on the identified bezels. One has a confirmed scratch; another bright example was accepted under the appearance method.

Is the scratch visible at every repeat seating under the candidate conditions? That's the part I'd want to hear next.

Hassan's question needs the actual expected presentation range, not deliberately identical placements that remove the variation you're investigating

Not consistently. Repeat captures show the scratch disappearing in one seating condition. We're reviewing that condition with the fixture and imaging people.

Can the trainer see that difficult example too? It would help explain why an apparently cleaner view isn't automatically better inspection.

Yes, both physical examples and their views are retained. Thanks Jamie, the operator is joining the holder review rather than just receiving its instructions afterwards.

Leave threshold tuning paused while that visibility gap is unresolved. You now have a concrete imaging problem, not a reason to relabel the scratch as acceptable

When a revision is proposed, compare it against both examples and the same presentation cases. A holder change should not silently narrow the inspection task to its easiest pose