A hospital is evaluated on how quickly it discharges patients. Discharge rates improve dramatically. So do 30-day readmission rates — patients are being sent home too soon, falling ill again, and returning.
The hospital optimised perfectly for the metric. And failed at the actual goal.
This is Goodhart’s Law.
What Is Goodhart’s Law?
Goodhart’s Law, formulated by economist Charles Goodhart in 1975, states: “When a measure becomes a target, it ceases to be a good measure.”
A metric is useful when it correlates with something that genuinely matters. But once you start optimising for the metric — once it becomes the goal rather than the proxy — people find ways to improve the number without improving the underlying reality. The metric and the thing it was meant to measure come apart.
Why It Happens
Metrics are always proxies. Discharge speed is a proxy for patient health outcomes. Test scores are a proxy for learning. Revenue is a proxy for value creation. When you reward the proxy directly, you optimise for the proxy — not the thing it was supposed to represent.
The mechanism is simple: any measurable metric can be gamed. And once the metric has consequences attached — bonuses, rankings, funding — people have strong incentives to game it.
Real-World Examples
Education. When teacher evaluations are tied to student test scores, teachers teach to the test. Test scores rise. Learning — of content not on the test — stagnates or falls. The metric goes up; the thing the metric was meant to capture doesn’t.
Social media. Platforms that optimise for engagement — likes, comments, shares — discover that outrage drives more engagement than insight. The engagement metric goes up. The quality of discourse goes down. Engagement was the proxy; it became the target.
Customer service. Call centres measured on call resolution speed find agents rushing callers off the phone before problems are actually resolved. Resolution speed improves. Customer satisfaction falls.
GDP. GDP measures economic activity. Activity that destroys value — cleaning up disasters, building things that immediately need replacing, producing goods nobody wants — counts as GDP growth. The measure captures something real but misses much of what matters for human wellbeing.
Academic citations. When academic prestige is measured by citation count, papers are written to maximise citations rather than to advance knowledge. Citation rings, self-citations, and papers that cite popular papers regardless of relevance proliferate.
The Cobra Effect Connection
Goodhart’s Law is closely related to the Cobra Effect — the phenomenon where a well-intentioned incentive produces perverse outcomes. The colonial government offered a bounty for dead cobras. People started breeding cobras to collect the bounty. The cobra population grew.
Both effects share the same root: when you attach rewards to a measure, people optimise for the measure, not the goal the measure was designed to track.
What to Do Instead
There is no complete solution. But several practices help:
- Use multiple metrics. Any single metric can be gamed. A portfolio of metrics measuring different facets of the goal is harder to game comprehensively.
- Rotate metrics regularly. Once people know which metrics matter, they optimise for them. Changing what you measure periodically reduces the gaming window.
- Keep asking whether the metric still tracks the goal. The relationship between a metric and the thing it measures erodes over time as people adapt. Regular audits of whether the number still means what it’s supposed to mean are necessary.
The Hardest Level in Mind Traps
Goodhart’s Law is Level 40 — the final level — in Mind Traps, a free quiz covering 40 psychology laws and cognitive biases. Players frequently confuse it with the Cobra Effect. The quiz scenario illustrates exactly why they’re related but distinct.
Leave a Reply