I find it weird that many businesses/organizations/people try to come up with one metric to keep things simple (and to be easy to optimize), but are then surpri

I find it weird that many businesses/organizations/people try to come up with one metric to keep things simple (and to be easy to optimize), but are then surprised by Goodhart's law. It's important to have multiple metrics that measure different things (and are orthogonal enough to make it hard to optimize them all) and then use judgement to decide when something has gone too far on one path. I look at ASICs which might have Ops/sec, Power, Cost, TTM (even those could be broken down to MACs/FLOPs, maxTemp, Design vs Unit cost, TikTok-cycle vs manhours). Perhaps the organization even wants to change direction, but the "one" sclerotic metric makes it hard to do that.