Goodhart's Law

A dashboard can make a complicated goal look manageable. Once a team chooses a number to represent success, meetings often turn toward improving that number, even when the real aim is broader.

A clinic tracks appointment duration to reduce waiting. If clinicians rush every visit to hit the target, the waiting metric may improve while patients leave with unanswered questions. The indicator needs a companion measure of whether care addressed the visit's purpose.

Goodhart's Law is commonly summarized as: when a measure becomes a target, it can cease to be a good measure. It is a warning about incentives and measurement, not a universal rule that every target fails.

Consider a help desk judged only by how quickly tickets close. Staff may resolve simple requests first, split difficult cases into several tickets, or close cases before users confirm a fix. The reported speed improves, while service quality may not.

The pattern does not require bad intentions. People adapt to what is rewarded, and a narrow metric gives them a clear signal about where effort will count. That adaptation can make the metric less representative of the outcome it was meant to track.

A useful measure can still guide attention. A response-time target may reveal delays, and a graduation rate can signal whether students are completing a program. Trouble begins when the indicator is treated as the goal instead of evidence about it.

A delivery company rewards drivers for the highest number of completed stops. A driver may prioritize nearby parcels and defer a complex delivery. The count remains accurate, but it no longer gives a complete picture of reliability across customers.

The risk rises when rewards or penalties depend heavily on one measure, when people can influence how it is recorded, or when the underlying goal is difficult to observe directly. In those settings, gaming is only one possible response; neglect is another.

Use a small set of indicators that reflect different dimensions of the goal. For customer support, pair speed with user-confirmed resolution and a sample of case quality, rather than assuming any one figure captures the whole experience.

Review how the metric changes behavior, not only whether the chart moves. Ask workers what they now do differently, and check whether those changes help the intended beneficiaries. Their answers can expose side effects invisible in the aggregate.

Targets should also be revisited. A measure that works during an early pilot can become distorted after it is tied to compensation, public rankings, or a quota. Keep room to revise it when the context or behavior changes.

The practical lesson is not to abandon measurement, but to treat a number as a partial signal. What would you want to inspect alongside a target before deciding that the underlying goal is improving?

Psychological perspective

From a behavioral perspective, targets focus attention and shape what people learn to repeat. This response is not evidence that workers are dishonest; people often follow the incentives and rules an organization has made most visible.

A metric is a model of an outcome, not the outcome itself. Its usefulness depends on context, feedback, and how people respond to it, so monitoring unintended effects is part of responsible measurement.

Thinking clearly

State the real-world outcome in plain language before choosing indicators. Then ask which parts are observable and which would be hidden if the score alone improved.

Pair the main metric with a quality or counterbalance measure, and periodically examine actual cases rather than relying only on averages. Choose companions that could reveal the most likely shortcuts or neglected groups.

After introducing a target, ask the people doing the work what behavior it rewards and what it makes harder. Adjust the target if that evidence shows the number is drifting away from the purpose.

When a score improves, what would help you tell whether the underlying outcome improved with it?