Goodhart’s Law

Information & Knowledge

Intermediate
Goodhart’s Law says that when a measure becomes a target, it stops being a good measure. It matters because metrics change behavior, and behavior optimized for the metric often drifts away from the real goal.
Difficulty
Intermediate
Time horizon
Medium
Risk sensitivity
High
Typical misuse
Using the law to reject all metrics instead of building better measurement and incentive systems

Core Idea

Definition

Goodhart’s Law is the principle that once people begin optimizing explicitly for a metric, that metric becomes less reliable as a proxy for the underlying phenomenon it was meant to measure.

In Plain English

A metric works best when it observes behavior. Once people start gaming it, the number no longer tells the truth as cleanly.

How It Works

Metrics are compressions of reality. They track one slice of performance so people can coordinate and evaluate progress. The trouble starts when the metric becomes the goal itself. People then optimize toward whatever raises the number, even if the underlying purpose is harmed. A school can raise test scores while weakening learning. A company can increase activity metrics while lowering true value. The model explains why measurement systems often decay under pressure: the act of rewarding the metric changes the system being measured.

When to Use

  • When designing incentives tied to performance measures
  • When a metric is improving but the real outcome feels worse
  • When evaluating dashboards, scorecards, or KPIs
  • When creating accountability systems for teams or institutions
  • When trying to prevent gaming, distortion, or proxy drift

Examples

Everyday

If you focus only on reading page counts, you may start skimming quickly for the number rather than reading deeply for understanding.

Professional

A support team measured on speed alone may respond fast while leaving customers unresolved, because the metric no longer tracks real service quality.

Extreme Case

A political or institutional system can become obsessed with reportable targets while the actual public outcome deteriorates behind the numbers.

Common Mistakes

  • Assuming a measurable proxy fully captures the real goal
  • Rewarding one metric so heavily that people sacrifice the underlying mission
  • Tracking activity because it is easy while neglecting quality because it is harder
  • Fixing distortion by adding more metrics without addressing incentive structure

Limits & Failure Modes

  • Not every metric becomes useless when targeted; some remain helpful if designed carefully
  • The law does not mean measurement is pointless
  • Overreacting can lead to vague goals with no accountability at all
  • Different metrics fail in different ways depending on the incentive environment

How to Practice

metric vs mission

For every important measure, name the underlying outcome it is supposed to proxy and check whether optimization is drifting away from that outcome.

gaming scan

Ask how someone could improve the metric while harming the real goal.

mixed indicator design

Use a blend of metrics, qualitative feedback, and spot checks so no single number carries the whole system.

Related Cognitive Biases

metric fixation

People mistake the proxy for the purpose and allow the number to replace the underlying aim.

salience bias

Visible, countable metrics crowd out harder-to-measure but more important realities.

incentive caused bias

Once rewards attach to a number, reasoning and behavior warp around preserving or improving it.

Related Mental Models

Related Skills

evaluating reliability
strategy definition
group dynamics mapping
detecting manipulation

Advanced Notes

Historical Origin

The principle is associated with economist Charles Goodhart and has broad relevance in policy, finance, education, and management.

Philosophical Context

It highlights the instability of proxies once they become objects of strategic optimization.

Further Reading

  • The Tyranny of Metrics by Jerry Z. Muller
  • Seeing Like a State by James C. Scott
  • Poor Charlie's Almanack by Charles T. Munger

Primary Domains

Management
Policy
Measurement