My sheets contain both restrained and released results, with some conditions unlabelled. Although the drawing specifies the dimension, the inspection procedure doesn't clearly define the supports.

Ask quality and the design owner to establish the intended acceptance condition, then group historical results by their documented measurement state to support comparison.

@VictorAllen0272 Same restraint label still doesn't mean comparable. Unknown supports and mixed fixture or process revisions could remain inside those groups.

That's a fair correction. State grouping addresses only one source of difference; the review also needs support conditions and fixture and process references, with missing information explicitly identified.

Should a fixture revision remain unknown when a colleague remembers a likely version but no record confirms it? We're disagreeing about whether to fill those gaps.

Mark it unknown. You can note the recollection separately as something to investigate, but don't turn a probable revision into a recorded fact.

On my bracket job, filling blanks from memory produced a very persuasive trend. Finding the original fixture sheets made the trend disappear. Not my favourite chart.

If the fixture holds it straight, why isn't that already the result we want?

Because the required inspection condition may be after release. Holding a shape under restraint doesn't establish the shape the assembly takes without that restraint.

I've had the drawing owner confirm the released inspection condition and its reference supports. Quality is adding those requirements to the sheet so the measurement state is explicit.

That establishes the intended inspection basis, but the measurement conditions of unlabelled historical results still remain unknown.

I'll preserve the uncertainty on those historical entries. Establishing the intended condition gives the review a basis without supplying missing facts about past measurements.

Where sample links exist, pair incoming fit-up observations with the released measurements. That allows the welding engineer to assess associations in the context of the documented process.

I've found varying incoming fit-up in the traceable records for assemblies that changed shape after release. The association is visible, although it doesn't establish causation.

I can describe the measurement uncertainty better, although the cause of released variation remains open. Any fixture change still needs a specialist assessment supported by the inspection basis and traceable records.

That fits the evidence. Clearer measurement uncertainty helps, but it doesn't yet explain the released variation or specify a fixture change.