Which QA metrics actually help vs which ones become theatre?

New VP asked for a "QA metrics dashboard". I've been in QA long enough to know that whatever I put on that dashboard is what my team will optimize for — for better or worse. Test count, pass rate, code coverage — these all feel like theatre. Escaped defects and MTTR feel real but are lagging. What metrics have you seen actually change behavior in a healthy way (i.e., drove better quality without perverse incentives)? And which ones did you regret putting on a dashboard?