Goodhart Law

Managers need metrics. Without them, organizations drift into anecdote, politics and intuition disguised as evidence. Metrics help leaders detect trends, allocate attention, evaluate progress and surface problems early. They are indispensable.

Yet metrics become dangerous when leaders forget what they are. A metric is not reality. It is a proxy for reality. It is a simplified signal that points toward something more important than itself.

Goodhart Law captures the problem with brutal clarity. In its popular form, it is usually stated as: "When a measure becomes a target, it ceases to be a good measure." The original insight from economist Charles Goodhart was slightly different but even sharper: "Any observed statistical regularity will tend to collapse once pressure is placed upon it for control purposes"1.

The management implication is profound. The moment a number determines reward, punishment, ranking, status, funding, or strategy, people begin adapting to the number. That adaptation often breaks the relationship between the metric and the underlying outcome it was supposed to represent.

Why metrics fail when targeted

A measure works because it correlates with something leaders care about. Revenue growth may correlate with value creation. Customer-satisfaction scores may correlate with loyalty. Defect rates may correlate with quality. Call resolution time may correlate with service efficiency. Story points may correlate with delivery throughput.

But correlation is not identity. The metric is not the thing itself.

Once a metric becomes a formal target, people naturally search for the easiest path to improve it. Sometimes that path also improves the real outcome. Often it does not. The problem is not bad intent alone. It is rational adaptation. If compensation, prestige, or survival depends on a number, people optimize the number.

This changes behavior and distorts the original signal. A call center measured heavily on short average handling time may push agents to end calls faster, even if customers need more help. A school judged on test scores may narrow teaching to the test. A software team measured on ticket closure may split work into smaller tickets or close low-value items first. A sales organization measured on quarterly bookings may discount heavily or sign poor-fit customers who later churn.

The number improves. The underlying reality may not.

Goodhart's original insight

Goodhart's observation emerged from monetary economics. Policymakers had identified statistical regularities in financial systems and attempted to use them as control targets. Goodhart argued that once a government relied on those regularities for policy control, the relationships would tend to break down because the system would adapt2.

That original context matters because it shows that Goodhart Law is not merely about cheating. It is about reflexivity. Systems change when participants know what is being measured and what consequences follow. The act of control changes the phenomenon being observed.

This is why the law applies far beyond economics. Modern organizations are full of feedback loops. Employees respond to performance reviews. Suppliers respond to procurement metrics. Students respond to grading systems. Engineers respond to delivery targets. Customers respond to loyalty programs. Algorithms respond to the data environments in which they are trained and optimized.

Once measurement enters the loop as a target, behavior changes. The metric is no longer a passive observer. It becomes an active force.

The proxy problem in management

Leaders rarely manage the ultimate goal directly. They cannot measure "value creation," "trust," "learning," "innovation quality," or "customer love" perfectly in real time. So they choose proxies.

That is unavoidable and often wise. The problem begins when the proxy is treated as sufficient.

A good metric should prompt questions, not end them. Yet many management systems do the reverse. Dashboards convert a complex reality into a small set of colored indicators. Incentives focus the organization on those indicators. Over time, the organization starts treating the indicators as the goal.

This is how performance systems become self-defeating. A hospital may reduce waiting-time metrics by redefining the waiting period. A consulting firm may improve utilization by keeping people busier on low-value work. A media business may optimize click-through rates while weakening trust and brand quality. A product organization may maximize engagement while degrading user well-being or long-term retention.

In every case, the metric began as a useful signal. It became damaging when it was elevated from indicator to objective.

The most common management mistake

The most common mistake is not using bad metrics. It is using single metrics too aggressively.

Single-number control is attractive because it simplifies accountability. Leaders want a few clear targets that align effort. Investors want comparable numbers. Boards want concise reporting. Employees want clarity. Software tools make dashboards easy to build. All of this pushes organizations toward simplification.

But simplification creates exposure. A single metric almost always leaves something important unmeasured. Once the metric becomes the basis of reward or judgment, people exploit the gap between the number and the underlying reality.

This is why Goodhart effects appear even in well-intentioned organizations. A team does not need to be unethical to game a metric. It only needs to notice that success is defined narrowly. Optimization then follows naturally.

The more pressure leaders place on a single number, the more likely it is that behavior will shift in ways that preserve the number while weakening the broader mission.

Examples across organizations

Goodhart Law is easy to recognize once leaders know what to look for.

In sales, a target for new accounts can encourage acquisition of low-fit customers. The bookings metric rises while profitability, retention, or support burden worsens.

In software, a target for deployment frequency can encourage smaller, lower-value releases or discourage necessary architectural work that reduces short-term output.

In recruiting, a target for time-to-fill can encourage speed over quality, creating expensive hiring mistakes.

In customer service, a target for low call duration can reduce true resolution quality.

In education, a target for test performance can shift teaching away from broader learning.

In consulting, a target for billable utilization can reduce knowledge-building, innovation and thoughtful client development.

In healthcare, targets can alter coding, triage, documentation and treatment behavior in ways that improve measured performance without necessarily improving patient outcomes3.

The pattern is consistent. The organization ends up serving the proxy.

Metrics can create moral drift

One reason Goodhart Law is dangerous is that it can gradually normalize behavior that people would otherwise question. When a metric dominates performance management, people start treating whatever improves the metric as legitimate by default.

Over time, teams stop asking whether the behavior advances the real goal. They ask only whether it will count.

This can create moral drift. Employees are not necessarily trying to deceive. They are trying to survive and succeed in the system leaders designed. If that system rewards visible numbers over actual value, the culture begins to separate appearance from substance.

This is why Goodhart Law is also a leadership ethics issue. A distorted metric does not only create operational inefficiency. It can erode integrity by teaching people to perform success rather than produce it.

Goodhart Law in digital businesses

Digital businesses are especially prone to Goodhart effects because they generate abundant measurable data. Everything can be counted:

clicks, sessions, retention curves, time on page, tickets resolved, velocity, churn, cost per acquisition, uptime, conversion, response time and engagement

This abundance creates both strength and temptation. Teams can learn rapidly from data, but they can also become imprisoned by what is easiest to count.

A product team that optimizes only engagement may encourage addictive behavior rather than meaningful value. A content team that optimizes only views may produce sensational material that damages trust. A growth team that optimizes only conversion may create onboarding that maximizes sign-ups but weakens long-term retention or customer fit.

The issue is not data richness itself. It is that digital environments make proxy optimization extremely efficient. Teams can run experiments, tune algorithms and redesign flows quickly. If the metric is wrong or incomplete, the organization can optimize itself into strategic error faster than before.

How to use metrics without being ruled by them

Goodhart Law does not mean leaders should distrust all measurement. It means measurement must be governed carefully.

The first safeguard is to define the underlying objective clearly. What is the organization truly trying to improve? Revenue quality, customer trust, learning depth, clinical outcomes, product value, resilience, long-term profitability, or something else? Without this clarity, the metric inevitably becomes the mission.

The second safeguard is to use metric sets rather than single metrics. Balanced measurement makes gaming harder because it forces trade-offs into view. A sales target should be paired with margin, retention, or customer quality. A service metric should be paired with resolution quality and satisfaction. A software delivery metric should be paired with reliability, customer impact and technical-health indicators.

The third safeguard is to watch behavior, not only dashboards. If a number improves unexpectedly fast, leaders should ask what changed in behavior and whether the real outcome improved too. Goodhart effects are often visible first in practice, not in reporting.

The fourth safeguard is periodic metric review. A metric that was useful last year may be dysfunctional now. Organizations change, markets change and people learn how systems work. Metrics should therefore be treated as provisional tools rather than permanent truths.

Ask what the metric is hiding

A strong management habit is to pair every KPI with a shadow question.

If conversion rises, what happened to customer quality? If utilization rises, what happened to innovation and capability building? If churn falls, what happened to discounting or contract structure? If ticket closure rises, what happened to root-cause resolution? If test scores rise, what happened to broader learning? If deployment frequency rises, what happened to strategic architecture work?

This habit preserves judgment. It reminds leaders that performance systems contain blind spots. Every metric highlights something while obscuring something else.

The best leaders do not reject metrics because they are imperfect. They use them intelligently because they are imperfect.

From proxy worship to disciplined judgment

Goodhart Law is ultimately a warning against proxy worship. Organizations need numbers, but numbers are not self-interpreting. They are useful only when leaders remain clear about the purpose they are meant to serve.

A metric should illuminate performance, not replace thinking about performance. It should help leaders inquire, not permit them to stop inquiring. It should focus attention, not narrow reality until the institution serves the dashboard more faithfully than the mission.

This is why the strongest organizations combine quantitative discipline with qualitative judgment. They understand that metrics are essential but incomplete. They build balanced scorecards, review unintended consequences, invite challenge and remain willing to retire measures that no longer serve the real goal.

Goodhart Law does not tell leaders to ignore numbers. It tells them to remember what the numbers are for. Once that is forgotten, the measure stops guiding value and starts distorting it.

• 1Goodhart's law and its original formulation
• 2Goodhart's Law: recognizing and mitigating the manipulation of measures
• 3When a measure becomes a target