Batch Four, Held Back Nine Days

Objectives

• Complete the batch to the same specification as the three before it.
• Verify the specification itself, not only conformance to it.
• Release nothing while a question about the measurement remained open.

Production Details

The batch ran normally. Cutting, assembly, and finishing all completed within the expected window, and no station reported an issue during the run. The batch was staged for release on schedule.

The hold came from a comparison that is not part of the standard check: batch four's finished measurements were placed beside batch one's. Every piece in both batches was inside tolerance. The batch four pieces were clustered near one edge of that tolerance and batch one's were centred.

Quality Inspection

Individual inspection passed at every station. The drift was invisible at the level of a single piece and only appeared when the batches were compared as populations — which is the argument for keeping the earlier measurements rather than discarding them once a batch ships.

Issues

• Finished measurements clustered at one edge of tolerance rather than centred.
• No single station responsible; the drift accumulated across three of them.
• The standard check could not have caught this, by design — it tests pieces, not distributions.

Solutions

The three stations were re-referenced against the master pattern rather than against their own most recent output, which is how the drift had propagated. The batch was re-measured after re-referencing and released once the distribution recentred.

Production Statistics