Our bracket baseline changes when the checking fixture is changed

Both have drawings. The earlier checking fixture was retained for another bracket variant, and its label is similar to the current one. Quality is comparing the datum arrangement, not only the label.

Were the physical brackets identified against the correct variant too, Ada, or could the sample itself have moved between those two jobs?

I would hold the claimed movement out of the estimate while that is checked. Your separated stages remain useful, but a precise pair of readings can still compare different dimensions.

Isaac, the retained bracket and traveller identify the current variant. Quality found the first fixture checks from a different reference face. Those two readings cannot measure a change on this bracket.

Then keep the historical readings with their own references and replace the invalid comparison. Can quality check the retained bracket using the current method?

They did. Retained bracket passes the current released check. The earlier reported failure used the wrong checking fixture, not an observed movement between comparable measurements.

That resolves this example. I would still leave the correction history of the other brackets as recorded; one mistaken check does not establish that the whole correction workload was unnecessary.

Agreed. This bracket's invalid comparison is corrected in the baseline, while the other linked correction observations remain. Quality has separated the two checking fixtures and clarified the variant labels at the station.

Thanks for coming back with the fixture finding. The useful question was what each tool checked, not which one looked older.