Which blank measurements would explain our changing stick-out?

They're numbered and retained, fortunately. The sheet matches those numbers; it just leaves out most of what we'd need to explain the readings.

Keep that identity through the next check. Ask quality and the setter to agree overall length, exposed-length reference and the seating-end inspection on one sheet.

And measure a repeat on the same blank, so you can see whether the seating result changes without changing the part.

Record repeated loadings as separate observations under the same blank identity; overwriting the first reading with the better-looking second would destroy the comparison.

Dan, yes. 'Final reading' is not enough when the question is repeatability.

Zara, two setters contributed. We found they measured exposed length from different faces on the chuck. The revised sheet has the agreed reference marked; their original readings remain separate.

Ah. That could explain a chunk of the spread before any loading fault enters the picture. Are they both using the revised method for the repeat?

Yes. Quality watched them use the agreed reference. The apparent difference between setters disappeared, but two blanks are genuinely longer overall and still protrude further.

Did the same-blank repeats agree as well?

For most. One blank gives an occasional larger reading on repeat, and the setter says it sometimes feels different at the stop. That observation is tied to its number now.

Keep that blank aside for the end inspection. Do not average the odd readings into the group and call it a longer part.

Farah, include the setter's observation in the record as an observation, not a confirmed seating failure; it is useful without choosing the cause yet.

Quality found a damaged seating edge on that blank and excluded it from the study set. We have not checked whether that damage accounts for every odd repeat. The other identified results reconcile with the overall lengths.

Thanks for separating those. Does the study need a replacement sample, or do you have enough accepted blanks covering the ordinary range?

Also worth keeping the damaged blank available for the people investigating it, rather than letting the exclusion mean it vanishes into scrap before they've finished.

Retained for inspection, Luca. Zara, quality selected another accepted blank from the same length group, and its repeated measurements agree. That gives the proposal a usable baseline while the damaged piece remains a separate question.

Who will teach the shared measurement reference to the relief setter?

And keep the original two-reference sheet as an example in that handover; the numbers looked plausible until somebody asked where they started.