Goodhart’s Law, explained.
Goodhart's Law describes how a measure can lose its usefulness as a proxy when people are rewarded for optimizing the measure itself.
Why it happens
A metric can track quality before becoming a target. Once incentives change, people may improve the number through shortcuts that do not improve the underlying goal. The important question is how behavior adapts.
A measure can lose its connection to the goal it once tracked when people begin optimizing it. Incentives, gaming, and changes in behavior can all weaken the relationship.
Look for this pattern
Keep the goal and the proxy separate. Evidence of a higher reported score is not enough; look for a corresponding improvement in the outcome the metric was meant to represent.
A worked example
A support team's speed target
A company uses average ticket-closing time as a rough measure of responsiveness.
If closing tickets quickly becomes the only rewarded behavior, staff may close unresolved cases or avoid difficult ones.
Pair speed with resolution quality and follow-up outcomes so the incentive reflects the actual service goal.
Where this idea is useful
A target should be evaluated alongside the underlying purpose it is meant to serve.
A common misconception
“All targets inevitably fail.”
The risk depends on incentives, the proxy's connection to the goal, and the available ways to game it. Some targets remain useful.
What this explanation leaves out
- Targets do not always fail. The effect depends on the design of the measure, the incentives, and how easily behavior can adapt.
How is Goodhart's Law different from ordinary measurement error?
Measurement error can exist before a target is introduced. Goodhart-style problems emphasize how optimizing the target changes behavior and can weaken the relationship between the measure and the objective.
How could someone improve your favorite metric without improving the thing you care about?
Associated thinkers
Associations marked provisional are awaiting source review.