Goodhart's Law
The moment a measurement becomes a target, it stops being useful — because people start optimising the number instead of the thing it was supposed to represent.
Why PMs should care
Any metric you turn into an OKR becomes the thing the team games — not the thing you originally wanted to improve. Proxy metrics are useful at first: they're cheaper, faster, and easier to read than the real outcome you care about. But the moment you attach compensation or headcount to one, it starts to quietly rot.
The failure isn't about dishonesty. People are very good at finding the shortest route to a number. And once the number is on a dashboard, 'what the number was originally supposed to represent' isn't really anyone's job anymore.
The defence is to pair any target metric with a counter-metric — something that should get worse if the target is being gamed. And every year or so, change how the target is defined, before anyone gets too comfortable working around it.
Example in product work
A support team is rewarded on 'tickets closed per day'. Quarter-on-quarter, tickets closed climbs 35%. Customer-reported resolution — the thing the metric was meant to proxy — drops 12%.
Investigation reveals agents have learned to close tickets with a templated 'thanks for reaching out, please let us know if you need further help', which resets the SLA clock, marks the ticket resolved, and transfers the problem to the next agent when the customer re-opens.
The metric climbed. The actual work got worse. And the agents weren't being dishonest — they were being rational inside the incentive they were handed.
What to do when you see it
- Proxy metrics work well until you tie compensation or headcount to them. Then they quietly go bad.
- The problem isn't dishonesty. Teams are very good at finding the shortest path to a number.
- Always pair a target metric with a counter-metric — something that should drop if people are gaming the target.
- Rotate the definition of key metrics every year or two, before anyone gets too comfortable gaming them.
Sources & further reading
- Problems of Monetary Management — Goodhart, 1975The original paper where Goodhart articulated the observation that became his eponymous law.
- Goodhart's Law: Are Academic Metrics Being Gamed? — Manheim & GarrabrantA modern formalisation of Goodhart's Law with four distinct failure modes.