Goodhart's law
Goodhart's law states that when a measure becomes a target, it ceases to be a good measure. Once people are rewarded for a metric, they optimise the metric itself — often at the expense of the real goal it was meant to track.
✦ Widely referenced — cross-referenced 6× across this reference (1 related ideas · 4 hubs) · The State of Thinking 2026 →
How it works
Whenever you set a target or KPI, distinguish the metric from the underlying goal it stands for. Then ask how people could hit the number without achieving the goal — and assume they eventually will, because incentives reward the measure, not the intention.
The moment a number controls rewards, it starts measuring people's ingenuity at hitting numbers.
How to use it
- Designing metrics and targets that are hard to game, or pairing them with counter-metrics.
- Diagnosing why a once-useful KPI has stopped reflecting reality.
- Staying sceptical of any system optimised toward a single number.
Worked example
A hospital judged on ER wait times starts logging patients as "seen" the moment a nurse greets them, or keeps ambulances queued outside. The wait-time metric improves while actual care doesn’t — the target corrupted the measure.
Where it fails
The lesson isn’t "never measure" — you need metrics to manage. The error is trusting any single metric too much. Over-correcting into metric-phobia leaves you flying blind; the fix is better, harder-to-game measures, not none.
- Gaming takes time to develop; a fresh metric can be a faithful target for a while, and the law says little about how fast it corrupts.
- Some measures resist gaming because they are the goal — profit for an owner, survival for an organism — so the law applies unevenly.
- The law names the failure but not the remedy; paired metrics, rotation, and human judgment are separate design work it does not do for you.
The counter-model: Skin in the game — Goodhart shows proxies corrupt when targeted; skin in the game aligns actors with real outcomes instead of measured stand-ins.
How to apply it, step by step
- List the metrics people in the system are rewarded or punished on.
- For each, ask what behavior maximizes the number without serving the real goal.
- Watch for divergence: the metric improving while ground-truth signals stagnate.
- Pair each target with a counter-metric that gaming would visibly degrade.
- Rotate or retire a metric once the gap between number and goal appears.
The deeper point
It is the reason almost every dashboard eventually lies. The moment a number controls rewards, it starts measuring people’s ingenuity at hitting numbers — which is why the most-optimised metric in any organisation is usually the least trustworthy.
Frequently asked
- What is Goodhart's law?
- It states that when a measure becomes a target, it stops being a good measure. Once people are rewarded for a metric, they optimise the metric itself — often undermining the real goal it was meant to capture.
- What is an example of Goodhart's law?
- Paying programmers per line of code produces bloated, padded code. The metric (lines) is hit while the goal (good software) suffers, because the reward is attached to the measure, not the outcome.
- How do you avoid Goodhart's law?
- Use multiple metrics that are hard to game together, pair targets with counter-metrics, and keep checking whether the number still reflects the underlying goal. Treat any single optimised metric with suspicion.
Related
Keep reading
Plato
Almost every later debate about reality, justice, or beauty begins as a footnote to him.
See this alongside the other thinking tools of product management, leadership & managing people, hiring and software engineers.
The books behind better thinking
Listen to any of these free. Start a free Audible trial and get your first audiobook on the house.
Prefer to read? The canonical picks:
- Thinking, Fast and Slow — Daniel Kahneman
- The Art of Thinking Clearly — Rolf Dobelli
- The Great Mental Models, Volume 1 — Shane Parrish
- Poor Charlie’s Almanack — Charlie Munger
- Super Thinking — Gabriel Weinberg & Lauren McCann
- Seeking Wisdom — Peter Bevelin
As an Amazon Associate, ReadGlobe earns from qualifying purchases and Audible trials — at no extra cost to you.
Cite this page
ReadGlobe. (2026). Goodhart's law. https://readglobe.com/model/goodharts-law/
"Goodhart's law." ReadGlobe, 29 May 2026, readglobe.com/model/goodharts-law/.
Primary source: Wikipedia
Editorial synthesis © ReadGlobe 2026, drawing on the mental-models tradition (Charlie Munger, Farnam Street) and the primary sources for each model. · Last reviewed 2026-05-29.