A service rewards staff for closing cases quickly. Some learn to resolve straightforward problems faster; others learn to close difficult cases before the person’s problem is actually solved. The same rising score can contain real improvement and strategic avoidance.

Failures of proxy optimization include deliberate gaming, but also less intentional effects. A measure may be noisy, work only within familiar conditions, or cease to represent the goal once people reorganize their behavior around it.

The number becomes part of the situation

Before becoming a target, a measure may summarize behavior. After becoming a target, it also helps cause behavior. People decide what to prioritize, which cases to accept, and what to record in response to its consequences.

Adding more measures can reveal some distortions, but it can also increase reporting work and produce a more complicated game. Independent observation, attention to neglected outcomes, and opportunities for people affected by the service to describe their experience can provide different kinds of evidence.

The important distinction is between the score and the reason for caring about it. A useful measure remains a tool for understanding the work. It becomes dangerous when improving the number is treated as sufficient proof that the work has improved.

Real cases