Evaluation Ownership: who answers when the model improves output but worsens the decision

Problema In many teams, evaluation improves while the decision gets worse. Model metrics go up, output looks cleaner, and the lab reports progress. At the same time, overrides, rework, and human escal