Every metric creates a shadow behaviour.
The moment an organisation announces that something will be measured, people begin adjusting their work around it. This is often the point of the metric. We measure sales because we want revenue to grow, response time because we want customers to receive help quickly and product delivery because we want the company to keep improving.
But every measurement illuminates one part of performance while leaving another part in the shadows. People optimise what is visible, and the organisation may quietly damage whatever the metric fails to capture.
I like to measure things. I prefer to see performance in black and white because a business cannot be managed well through impressions alone, especially as it grows. When a team is small, the founder may understand what everyone is doing through daily interaction. As the organisation expands, that informal understanding becomes unreliable, and clear measurements are needed to show whether work is producing the intended result.
Yet the solution is not to measure everything without thinking. A badly designed metric can produce behaviour that looks successful on a dashboard while weakening the company underneath it.
People study the measurement system
Metrics work in a way that resembles preparing for an examination. A student who wants to perform exceptionally well does not merely hope to pass. The student studies the syllabus, understands what the examination is likely to test, develops a routine and allocates time according to the importance of each subject.
Employees do something similar. Once they know how performance will be evaluated, they learn the organisational syllabus. They direct attention towards the activities that improve the score and reduce attention elsewhere.
This is not necessarily dishonest. Rational people respond to incentives. If a customer-service representative is evaluated mainly on how quickly tickets are closed, the person will naturally attempt to close more tickets. The unintended consequence may be that difficult cases are closed before customers feel that the underlying problem has been resolved.
If a sales employee is judged by the number of calls made, the call volume may rise while the quality of conversations falls. If an engineer is assessed primarily by the number of features released, the company may receive more features alongside more instability and technical debt. If a recruiter is rewarded for how quickly vacancies are filled, the organisation may hire faster while reducing the quality of the people it brings inside.
In each example, the metric works. It changes behaviour. The problem is that it changes more behaviour than leadership intended.
The reasoning behind a metric matters
When someone presents a metric, I do not want to see only the number. I want to understand how the person arrived at it, what the measurement is intended to reveal and why it should predict the outcome the company cares about.
The reasoning allows me to assess the quality of the metric. It also helps me see what may have been deliberately or unintentionally excluded.
Suppose a team proposes measuring the number of new customers acquired. That appears sensible, but the reasoning should continue. Are these customers paying? Do they remain active? What does it cost to acquire them? Are they the kind of customers the product can serve well? A campaign can produce many registrations while attracting people who never complete onboarding or generate enough revenue to justify the acquisition cost.
The headline metric may be positive while the business outcome is poor.
This is why a metric should be connected to a theory. We are measuring this behaviour because we believe it will lead to this intermediate result, which should contribute to this business outcome. When the theory is visible, leadership can test each link rather than assuming that movement in one number represents progress everywhere.
A metric without an explanation can become a ritual. A metric with a clear causal argument becomes a hypothesis the business can examine.
What is missing may matter more
Every performance measure should be accompanied by a second question: what important behaviour does this metric fail to capture?
Revenue does not necessarily capture profitability. Speed does not necessarily capture accuracy. Customer acquisition does not capture retention. The completion of an individual objective may not capture the support another team needed. Uptime does not show whether an important workflow is confusing, and customer satisfaction after a support interaction may not reveal customers who became so disappointed that they left without complaining.
The omitted behaviour becomes especially important when improving the measured number makes it easier to damage something else. A company that rewards revenue without examining the quality of revenue may encourage excessive discounts or contracts the operations team cannot fulfil. A company that rewards cost reduction without measuring service quality may become more efficient by transferring inconvenience to customers.
These are shadow behaviours. They live outside the primary measurement but are produced by it.
Leaders should look intentionally for them. If this number becomes the most important target, what shortcut will become attractive? What responsibility might people ignore? How could someone achieve an excellent score while making the company worse?
Designing the metric from the perspective of how it could be manipulated often reveals weaknesses that ordinary planning misses.
Metrics quietly build culture
Measurements do not remain inside performance reports. They tell employees what kind of company they are working for.
If individual targets matter more than shared outcomes, the culture will become less collaborative. If mistakes are counted without distinguishing thoughtful experiments from carelessness, employees will become more conservative. If only visible output receives recognition, people will neglect preventive work whose success is measured by problems that never occur.
The culture may change even though nobody explicitly intended to change it. The company simply repeats the measurement until people internalise the behaviour required to perform well within it.
This is why I want to understand not only what a metric measures, but what kind of organisation it is likely to create. A metric can silently sponsor a culture the founder would never approve if it were described openly.
For instance, telling leaders to own departmental OKRs may appear to create accountability. But if one team achieves its result by withholding help from another team whose success depends on it, the system has rewarded fragmentation. The person has optimised the visible score while damaging the company’s collective outcome.
A better system must account for dependencies. If your work materially affects another team’s deliverable, your performance should include whether that dependency was fulfilled. The metric should reward people for helping the organisation win, not merely for producing an attractive personal report.
Use balancing measures
One way to reduce shadow behaviour is to pair an important metric with a balancing measure. The purpose is not to create an enormous dashboard, but to ensure that progress in one area is not purchased through unacceptable damage elsewhere.
Measure sales alongside margin or collection. Measure customer acquisition alongside activation and retention. Measure support speed alongside resolution quality and repeated complaints. Measure product delivery alongside reliability and customer adoption. Measure operational efficiency alongside error rates and customer impact.
The balancing measure should reflect the most likely unintended consequence. If the company knows that speed can reduce accuracy, both should remain visible. If aggressive growth can weaken cash flow, leadership should examine them together.
However, adding more numbers is not always the answer. Too many metrics can make priorities unclear and give people enough data to select whichever figure presents their work favourably. A company needs a small number of measures that collectively describe the result, the health of the process and any critical constraint.
The objective is not perfect measurement. Human work is too complex for every valuable behaviour to be captured in a spreadsheet. The objective is to make the most important trade-offs visible and to prevent one number from becoming a substitute for judgment.
Review the behaviour, not only the score
Leaders should periodically examine how people are responding to the measurement system. What has changed since the metric was introduced? Are employees making better decisions, or merely becoming better at presenting the number? Have new bottlenecks appeared? Are customers, colleagues or other teams carrying an unexpected cost?
Qualitative evidence matters here. Customer conversations, operational reviews and direct observation can reveal what the dashboard cannot. A number may show that response time has improved, while listening to customers reveals that they are receiving quick but unhelpful answers.
The company should also be willing to revise a metric. A measurement that was useful at one stage may become distorting after people learn to optimise it or after the business model changes. Changing it is not an admission that measurement has failed; it is evidence that leadership is still paying attention to the behaviour being created.
Measure with humility
Metrics are essential because they force a business to move beyond anecdotes. They allow leaders to compare expectation with reality, identify patterns and hold people accountable as the organisation becomes too large to manage through personal observation.
But numbers do not eliminate the need for thought. Every metric is an incomplete representation of a more complicated reality. It should clarify judgment rather than replace it.
Before making a measurement important, ask how it was constructed, what outcome it is expected to predict and what it leaves outside the frame. Ask how a rational employee might maximise it, and whether the resulting behaviour would strengthen or weaken the company.
Then continue watching. The true effect of a metric is not the number appearing in the report. It is the behaviour spreading through the organisation because people know that number matters.
Measure what is important, but also examine the shadow it casts. The company is being shaped by both.
Leave a comment