Campbell's Law

Leaders rely on quantitative indicators because organizations are too complex to manage by intuition alone. Metrics bring visibility. They help compare performance across teams, identify problems early, allocate resources and create accountability. Without them, many organizations drift into anecdote, politics and selective storytelling.

Yet metrics become dangerous when they become powerful. The more a number determines funding, ranking, reward, punishment, reputation, or survival, the more pressure people feel to improve that number by any available means.

That is the core of Campbell's Law. Donald T. Campbell's original formulation remains one of the sharpest warnings in performance management: "The more any quantitative social indicator is used for social decision-making, the more subject it will be to corruption pressures and the more apt it will be to distort and corrupt the social processes it is intended to monitor"1.

This is not a narrow research-methodology observation. It is a management law. It explains why schools teach to the test, why service teams optimize call duration instead of resolution quality, why utilization targets can degrade consulting capability and why companies sometimes end up serving the dashboard instead of the mission.

Why indicators become corruptible

A quantitative indicator is useful because it captures some aspect of a reality that leaders care about. Test scores may reflect learning. Customer-satisfaction scores may reflect service quality. Sales quotas may reflect commercial performance. Utilization may reflect productive deployment of talent. Ticket-closure counts may reflect throughput.

But an indicator is not identical to the underlying reality. It is only a proxy.

Once leaders attach consequences to the proxy, they create pressure to exploit the gap between the number and the thing the number is meant to represent. That gap is where corruption begins.

Sometimes the corruption is direct. People may manipulate definitions, delay reporting, exclude inconvenient cases, or simply falsify data. Other times it is subtler. Teams may narrow the activity so it better fits the measure, even when that weakens the broader purpose. A teacher may focus heavily on test formats. A sales team may chase easy bookings instead of high-quality customers. A support team may reduce call length while leaving problems unresolved.

In both cases, the indicator improves while the underlying system becomes less healthy.

Corruption of the indicator and corruption of the process

One of the most useful aspects of Campbell's Law is that it highlights two kinds of corruption.

The first is corruption of the indicator itself. This occurs when the number is manipulated without real improvement in the underlying outcome. Examples include reclassifying cases, excluding difficult instances from measurement, shifting timing, or coaching behavior specifically to score well on the metric.

The second is corruption of the social process. This occurs when people reorganize work around what the indicator captures and neglect what it ignores. A school narrows the curriculum to improve tested performance. A hospital prioritizes measured treatment speed while overlooking less visible dimensions of patient care. A software team optimizes deploy counts while deferring architectural integrity.

This distinction matters because many leaders look only for obvious manipulation. They ask whether the data is true. Campbell's Law requires a deeper question: even if the number is accurate, has the activity itself been distorted in order to produce it?

That is the more serious danger. A system can tell the truth numerically while becoming less truthful operationally.

Why high-stakes metrics are especially risky

Not every metric creates the same level of distortion. The danger rises when a number becomes high stakes. High-stakes metrics influence compensation, promotion, funding, ranking, compliance outcomes, vendor selection, public reporting, or executive status.

As stakes rise, behavior adapts more aggressively. People spend more time optimizing the indicator, learning its weaknesses and protecting themselves from downside consequences. This is not necessarily unethical behavior. Often it is rational adaptation to the rules leaders have set.

That is why Campbell's Law is especially relevant in modern performance systems. Organizations now have more data, more dashboards and more algorithmic performance tools than ever before. The temptation is to believe that more measurement creates more control. In reality, more measurement can create more gaming if leaders do not understand how incentives reshape behavior.

The indicator becomes part of the system. It no longer observes the process from outside. It changes the process from within.

The education example remains the clearest

Campbell himself discussed achievement testing as a clear illustration. Under normal teaching conditions aimed at broad competence, test scores may provide useful information. But when test scores become the dominant goal of the teaching process, they lose value as indicators and distort education in undesirable ways2.

The same dynamic has been documented in research on high-stakes testing. Evidence compiled in education settings includes administrator and teacher cheating, student cheating, exclusion of low-performing students from testing, misrepresentation of dropouts, narrowing of curriculum, teaching to the test and score-reporting problems3.

This example matters because it is easy to generalize. The test score is not education. It is a partial measurement of education. Once it becomes the dominant target, the institution adapts to the test rather than to learning. The number survives. The purpose degrades.

Business leaders make the same mistake whenever they confuse a measurable slice of value with the total value they want the organization to create.

Campbell's Law in business

Business systems are full of vulnerable indicators. Revenue growth, margin, cost reduction, sales pipeline, NPS, utilization, churn, call duration, story points, deployment frequency, defect counts, engagement metrics and productivity dashboards can all be useful. They can also be corrupted.

A sales target tied heavily to bookings may increase low-quality deals that churn later. A utilization target in professional services may reduce time for training, innovation, knowledge building and relationship development. An NPS target may encourage survey manipulation or selective sampling. A service-level target may improve reported response time while leaving root problems unsolved. A software-delivery metric may encourage superficial releases instead of meaningful product progress.

The danger grows when leaders praise the metric more consistently than the mission. People quickly learn what counts. Once they know that, they begin managing for visibility rather than value.

Campbell's Law therefore explains why some businesses become strangely efficient at producing numbers that do not correspond to durable performance. The system has learned how to look healthy without becoming healthier.

Campbell's Law and Goodhart's Law

Campbell's Law is closely related to Goodhart's Law, but the emphasis is slightly different. Goodhart's Law is commonly summarized as: when a measure becomes a target, it ceases to be a good measure. Campbell's Law is more explicit about corruption pressure and about the distortion of the underlying social process.

This distinction is useful in practice. Goodhart's Law alerts leaders to proxy failure. Campbell's Law reminds them that metric pressure can damage the very activity they are trying to improve.

In business, both laws often operate together. A target becomes dominant. People respond. The measure becomes less informative. The work behind it changes shape. Leaders then intensify control because the indicator no longer behaves as expected, which creates even more distortion.

That cycle is difficult to stop once it becomes embedded in dashboards, compensation plans, public commitments and performance reviews.

The ethical dimension

Campbell's Law is not only an operational warning. It is also an ethical one. A metric-driven culture can slowly redefine integrity if employees learn that appearance matters more than substance.

People may not think of themselves as gaming the system. They may think they are being responsive, pragmatic and performance-oriented. They are giving leadership what leadership appears to want. But if leadership wants the number more than the purpose, the organization begins to reward representational success over real success.

That moral drift is dangerous because it is gradual. Small compromises become normal. Definitions become flexible. Untargeted work receives less care. Qualitative judgment is dismissed as softness. Eventually the institution loses contact with what the metric was supposed to help protect.

This is one reason Deming warned against the myth that only what can be measured can be managed4. Management requires attention to what matters, including aspects that are harder to quantify cleanly.

How leaders can reduce distortion

The answer is not to stop measuring. It is to measure with humility.

The first step is to define the true objective. What is the organization really trying to create or protect? Learning, customer trust, patient well-being, product value, resilient revenue, service quality, long-term capability, or something else? If this is vague, the metric will inevitably take over.

The second step is to use multiple indicators. A single number is easier to game than a balanced set. Sales should be viewed with retention and margin. Service speed should be paired with resolution quality. Utilization should be paired with capability building and client outcomes. Delivery speed should be paired with reliability and user value.

The third step is to inspect behavior around the number. If a metric improves sharply, leaders should ask what changed in the underlying process, not only whether the score is higher. Audit for exclusions, timing shifts, boundary changes, reclassification, selective sampling and neglected adjacent work.

The fourth step is to preserve qualitative judgment. Numbers help focus attention, but they cannot fully capture trust, judgment, learning, ethics, creativity and the quality of human interaction. Leaders should deliberately include narrative review, case-level analysis and expert judgment where the stakes justify it.

The fifth step is to revisit metrics regularly. A useful indicator can become dangerous once people learn how to optimize around it. Metrics should therefore be treated as living tools rather than permanent truths.

Ask what the score is costing

A practical discipline is to pair every important metric with a cost-of-optimization question.

If bookings are up, what happened to customer quality? If response time improved, what happened to resolution depth? If utilization rose, what happened to learning and innovation? If defect counts fell, what happened to reporting behavior? If test scores increased, what happened to broader competence? If satisfaction scores rose, what happened to sample integrity?

These questions help leaders see side effects before they become system characteristics. They also signal to teams that leadership is paying attention to purpose, not just to the scoreboard.

That changes behavior. When people know the organization examines how results are achieved, they are less likely to optimize the visible indicator at the expense of the underlying mission.

Protect the process, not only the number

Campbell's Law is ultimately a warning about substitution. The indicator was created to help leaders see reality more clearly. Under pressure, the indicator can become a replacement for reality.

That substitution is expensive because it distorts both information and behavior. Leaders believe they are managing more scientifically, while the system is quietly adapting to produce compliant numbers. The organization becomes easier to score and harder to understand.

The real leadership task is therefore not merely to choose metrics. It is to protect the purpose behind the metrics. That requires multiple measures, contextual judgment, regular review and a willingness to ask whether the system is becoming better or only more legible.

A good indicator should illuminate performance. It should not become the thing the organization serves. Once that happens, the signal is no longer being protected. It is being consumed.

• 1Campbell's original formulation
• 2Campbell on achievement tests
• 3The inevitable corruption of indicators
• 4NN/g on Campbell's Law and Deming