A pernicious product management pattern is data-driven feature testing - every single possibility is A/B tested to death. Even if they get the cohorting done co

A pernicious product management pattern is data-driven feature testing - every single possibility is A/B tested to death. Even if they get the cohorting done correctly so that you get a "stable" experience for that feature, the bar for changing the UX is much lower than it has ever been. There's no direct metric capturing UX stability, and I've never seen a real attempt at measuring what rate of UX change is sustainable. Bundled together with the PM career pattern of "add one big feature that has a measurable impact, then get a new job," we end up with an unstable stew where nothing works like it did last week.