Measurement is indispensable—and incomplete

Institutions need metrics.

Without measurement, performance can become anecdotal, accountability weakens and problems remain hidden.

But measurement becomes dangerous when what is measurable is treated as the whole of what matters.

Good systems measure what needs to be measured and preserve judgment for what cannot be responsibly reduced to a number.

A metric is a representation

A metric is not the reality itself.

It selects one feature of reality and expresses it in a form that can be compared.

This is useful precisely because it simplifies.

The simplification also creates limits.

The indicator-target problem

An indicator works best when it helps us understand a system.

Once the indicator becomes a high-stakes target, behavior adapts around it.

People may improve the underlying outcome.

They may also improve the indicator without improving the outcome.

What gets measured can crowd out what matters

If only measurable work receives institutional attention, less measurable work becomes vulnerable.

Trust, judgment, care, institutional memory, fairness and quality may be difficult to express through one number.

Their difficulty of measurement does not make them unimportant.

KPIs should have an ontology

Every KPI should answer a basic question: what exactly does this number represent?

Without that clarity, organizations begin treating proxies as reality.

A response-time metric measures speed.

It does not automatically measure whether the response was useful.

The Metric–Meaning Gap

I call the difference between a measured indicator and the underlying value the Metric–Meaning Gap.

The larger the gap, the more dangerous it becomes to govern directly from the metric.

Leaders should know where that gap is small and where it is substantial.

Measurement needs triangulation

A robust system rarely relies on one indicator.

Quantitative data can be combined with qualitative evidence, professional judgment, user experience and exception analysis.

Triangulation reduces the risk that one number becomes the institution’s entire field of vision.

The limits of ranking

Rankings create clarity.

They can also flatten important differences.

Two institutions with similar scores may serve different populations, operate under different constraints or produce different forms of value.

Context should travel with the number.

Accountability is larger than targets

An institution can hit targets and still fail its purpose.

Accountability therefore requires explanation, not just reporting.

What changed? Why? What trade-offs were made? What was missed?

Narrative reasoning remains part of responsible governance.

The Measure–Judge–Review Model

A mature governance cycle has three stages.

Measure: collect relevant evidence.

Judge: interpret the evidence in context.

Review: examine whether the measures themselves still represent what matters.

The third stage is often neglected.

Targets can distort professional judgment

When professionals know that only one number matters, they may prioritize what improves the number.

This can create tunnel vision.

Good systems protect space for professional judgment where cases differ materially.

Metrics should be proportionate to consequence

A low-stakes operational process can often be governed through simple quantitative measures.

High-stakes decisions involving rights, welfare or significant human consequences deserve broader evidence.

The higher the consequence, the less appropriate it is to rely on one metric alone.

The problem of measurable proxies

Institutions often measure what is available rather than what is ideal.

Attendance becomes a proxy for engagement. processing time becomes a proxy for service quality. complaint counts become a proxy for satisfaction.

Proxies are useful if their limitations remain visible.

They become dangerous when forgotten.

Unmeasured does not mean unmanaged

Some important values cannot be captured precisely.

Institutions can still govern them through standards, review, professional norms, qualitative evidence and independent scrutiny.

The absence of a single number does not require the absence of accountability.

The Quantification Boundary Test

Before turning a value into a target, ask:

Construct: What human or institutional reality are we trying to represent?

Proxy: How closely does the metric track that reality?

Distortion: What behavior might emerge if the measure becomes a target?

Exclusion: What important evidence will the number miss?

Review: How will we know when the metric stops being useful?

Data should increase visibility, not erase judgment

Digital systems can provide richer measurement than ever before.

The result should be better-informed judgment.

It should not be the elimination of judgment.

More data does not automatically produce more wisdom.

Human systems contain irreducible context

People do not arrive as standardized cases.

Their circumstances contain history, vulnerability, relationships and exceptions.

Institutional systems need enough flexibility to recognize material difference.

Good governance needs qualitative evidence

Interviews, case reviews, professional observations, user narratives and independent assessment can reveal what quantitative indicators miss.

Qualitative evidence should not be treated as inferior merely because it is harder to aggregate.

The dashboard boundary

A dashboard should tell leaders where to look.

It should not always tell them what to conclude.

The distinction protects institutions from governing by visualization alone.

The final governance principle

Measure enough to see.

Interpret enough to understand.

Question enough to know when the measure is distorting the purpose.

The most responsible institution is not the one with the most metrics. It is the one that knows what its metrics can see—and what they cannot.

Measurement can create false precision

Numbers often look more certain than the realities they represent.

A score may be calculated precisely even when the underlying construct is ambiguous.

Institutional confidence should reflect the quality of the construct, not merely the number of decimal places.

The audit of unintended behavior

Whenever a target is introduced, institutions should monitor the behaviors it creates.

Are staff avoiding difficult cases? Are users being redirected to protect a target? Are qualitative problems being hidden because they do not affect the score?

Metrics govern behavior whether designers intend it or not.

Targets should have sunset reviews

A useful metric can become obsolete as systems change.

Targets should therefore be reviewed periodically.

The question is not only whether the target was met.

It is whether the target still represents the purpose.

Institutional humility

Good governance includes the ability to say, ‘We do not yet have a reliable measure for this.’

Inventing a weak metric merely to create the appearance of control can be worse than acknowledging uncertainty.

Quantitative and qualitative evidence should challenge each other

Numbers can reveal patterns that stories miss.

Stories can reveal realities that averages hide.

Strong institutions use each form of evidence to test the other.

The final boundary

Measurement should illuminate human systems, not replace them.

The institution remains responsible for judgment even when the dashboard is sophisticated.