Goodhart's Law
Why it matters
The metric you choose decides what a team does all day. A dashboard can climb for months while revenue, retention or margin stay flat, and nobody can say why. The cause is rarely dishonesty. Most people simply respond sensibly to the incentive they were given.
It gets worse with scale. A small team knows what a number is for. A larger one inherits the number without the context, and a target set by one person is met by another who never heard the reasoning.
The cost shows up late. By the time revenue disappoints, the team has spent quarters tuning the wrong thing, and the data from that period no longer says much about the market.
How to apply it
- Prefer targets that are hard to inflate without producing the real result, such as closed revenue or retained clients rather than activity counts.
- Pair every target with a check metric that would expose gaming, such as quality alongside volume or close rate alongside leads.
- Watch an early-signal metric and an outcome metric together. When they disagree, trust the outcome.
- Retire or rotate a metric once the team has learned to hit it without moving the outcome.
- Before setting a target, ask what the cheapest way to hit it would be, and whether that would help the business.
What it is
Every metric is a stand-in for something harder to see. Calls booked stand in for interest, tickets closed for happy customers, leads logged for future revenue. As long as nobody is rewarded for the number, it tracks the real thing. Once it becomes a target, people find the cheapest way to hit it, and that route often skips the outcome.
The law is named after the British economist Charles Goodhart. The wording most people quote, that a measure stops being a good measure once it becomes a target, comes from the anthropologist Marilyn Strathern.
Common mistakes
- Rewarding activity. Calls made, emails sent and posts published are easy to inflate. Reward outcomes where you can.
- One metric alone. A single target invites a single shortcut. Pair it with a check.
- Never revisiting targets. A metric that worked last year may have been learnt and gamed by now.
- Blaming people. If the whole team games the number, the target is the problem.
- Hiding the reason. A team that knows what the number stands for can often spot when it stops doing so.
- Treating gaming only as a failure. That people found a shortcut tells you something about the incentive.
