Goodhart's law
When a measure becomes a target, it stops being a good measure.
What it means
Goodhart's law observes that once a metric is used as a goal or basis for control, people optimize the metric itself, eroding its connection to the underlying quality it was meant to track. The pressure invites gaming, narrow teaching-to-the-test behavior, and neglect of everything the indicator fails to capture. The mechanism is that any proxy is an imperfect stand-in for the real objective, and incentives exploit precisely the gap between proxy and goal. It is a foundational caution for performance management, algorithmic objectives, and any policy steered by a single number.
Examples
Rewarding call-center staff on calls-per-hour leads them to rush or hang up, improving the metric while degrading service.
Pay developers by lines of code and the code grows longer, not better. The metric climbs handsomely while the thing it stood in for — working software — quietly does not.
Judge hospitals on the share of patients dealt with inside four hours and attention shifts to the clock rather than the patient. The four-hour figure improves; the care behind it need not.
First described in Charles Goodhart (1975); sharpened by Marilyn Strathern (1997).