The question, and the paradox in the dashboard
Every organization measures. Revenue, satisfaction, throughput, quality, engagement - the metrics proliferate, the dashboards grow, the review cadences accelerate. And in many of these organizations, something paradoxical is quietly happening: the more precisely they measure, the less accurately they understand. The numbers are correct. The behavior they produce is not the behavior anyone designed the system to encourage.
The explanation is not that measurement is a bad idea. It is that two entirely different uses of the same instrument have been collapsed into one - and the collapse is invisible precisely because the tools look identical from the outside.
Two uses of the same instrument
Measurement is the act of making something legible in order to understand and improve it. It treats data as feedback - information that reveals the distance between where things are and where they need to be. Its opposite is blindness: operating without signal, unable to distinguish what is working from what is not.
Control is the use of metrics to evaluate, pressure, and surveil. It treats data as evidence - information that reveals something about the people being measured: their effort, their honesty, their compliance. Its opposite is trust: the willingness to use data as information without weaponizing it against the people who generated it. They always know which one it is. And they behave accordingly.
What control-oriented measurement actually produces
When measurement is experienced as control, the rational response is to optimize for the metric rather than for the underlying reality the metric was designed to represent. This is not dishonesty. It is intelligence applied to the wrong problem - the problem of surviving the review rather than delivering the outcome.
Three pathologies emerge reliably: local optimization (teams improve their metrics by moving the problem somewhere unmeasured), concealment (bad news travels slowly), and short-termism (the measured horizon becomes the planning horizon). The deepest damage is to the signal itself - the numbers are accurate but no longer telling the organization what they were designed to tell it.
Control produces the appearance of measurement.
Organizations learn to perform the metrics rather than move them.
The dashboard stays green. The understanding disappears.
What measurement in service of learning actually requires
Questions, not verdicts. Control-oriented review asks: why is this number not where we said it would be? Learning-oriented review asks: what does this number tell us about what we should do differently? The data is identical. The conversation it generates is entirely different.
Safety for bad news. The single most important condition for measurement to serve learning is that surfacing problems is experienced as valuable rather than dangerous. The response to the first bad number seen in a meeting sets the norm for every bad number that follows.
Metrics that stay connected to what matters. The discipline is asking, for every metric, what behavior it would rationally produce if someone wanted to look good without doing good. If that behavior is easy to execute and hard to detect, the metric is not measuring what it claims to.
The leader's real job: scientist, not auditor
An auditor checks whether the numbers are right. Their implicit assumption is that the people being measured are a potential source of error to be checked. A scientist asks what the numbers mean - what they reveal about the system, what they suggest about the next intervention. Their implicit assumption is that the data is a source of understanding. One produces compliance. The other produces learning.
The bottom line
The question is not whether to measure. The question is what the measurement is for. That question is answered by how leadership responds to bad news, whether honest assessment is visibly rewarded or subtly penalized, and whether the conversation around data is oriented toward explanation or exploration. The same numbers, in two different organizational contexts, produce two entirely different behaviors. One organization learns. The other performs the appearance of learning.
PEOPLE
Chapter 7 - Culture & Values