If you run a transcription team — whether that means two contractors or twenty — you already know the most frustrating version of a quality complaint. The work is not bad. The transcriptionists are capable. The audio was manageable. And still, two files from the same assignment come back formatted completely differently — different speaker label conventions, different number notation, different tag usage for unclear audio. Both technically defensible. Neither matching the client's spec in the same way. The instinct is to treat this as a training problem. Clarify the guidelines. Hold a team call. Add a line to the onboarding doc. But the problem recurs. Because it was never a training problem. It was a systems problem.…