# When the Number Becomes the Goal: The Measurement Trap in Public Life

> MKNIRALO — World — DEEP ESSAY

- Published: 2026-11-03
- Estimated reading time: 20 minutes
- Byline: Iona — International Institutions Observer
- Byline disclosure: Iona is a publication-owned house editorial byline, not a claim of an independently verified individual or a diplomatic correspondent with personal access. Articles use accessible documents and independent reporting, and avoid presenting disputed claims as settled facts.
- Author profile: https://mkniralo.com/authors/iona/
- Tags: Public Policy, Measurement, Governance, Public Services, Accountability, Education, Data Literacy, Institutional Design
- Cover image: https://mkniralo.com/images/measurement-trap-2026.svg

Institutions need performance numbers to make decisions and answer to the public. But when a proxy becomes a target, the number may improve while the underlying purpose remains unmet.

A school can enrol more children without helping each child learn. A hospital can process more appointments without improving every patient's experience. A government department can complete its annual list of reforms while a person seeking a basic service still encounters the same obstacles. These are illustrative possibilities, not evidence against a particular institution. They raise a question that reaches beyond any one country or profession: what happens when an easily counted measure is treated as the thing society really wants? Measurement is necessary for public accountability, but a number cannot always carry the weight of its underlying purpose. This essay develops a four-part way of interrogating that gap—Proxy, Incentive, Blind Spot, Verification. The framework is an original editorial diagnostic, not an empirically validated index or a claim of new field research.

## The seduction of an unambiguous number

A number has a special ability to travel. It can move from an internal spreadsheet to an agency report, from a budget hearing to a political speech, and from there into the public memory of whether an institution is succeeding. It appears comparable across time and space. A target can be given a line on a chart, a deadline and a colour. A complex public purpose rarely fits into the same small space.

That convenience solves a real problem. Most people cannot visit every school, observe every emergency department or audit every public office. Without indicators, a powerful institution can claim that everything is improving and leave citizens with little basis for challenging it. Numbers make performance discussable. They also help reveal problems that anecdotes hide, including differences between regions or groups.

The difficulty begins when a number selected as evidence is converted into a verdict. Imagine a fictional city announces that it has processed forty percent more housing applications. Perhaps the staff have become more capable and residents are receiving decisions faster. Or perhaps the count includes forms returned incomplete, duplicate requests or cases that were closed without solving a problem. The recorded activity may be accurate in both scenarios. What differs is the meaning attached to it.

The UK's National Audit Office has argued that meaningful performance measurement helps with accountability and learning, while its reviews also document the difficulties of selecting good measures and avoiding misleading incentives. The choice is not between numbers and judgement. The question is whether the numbers expose the real public purpose to scrutiny or let the institution replace that purpose with a more convenient description of itself.

Sources:
- [Measuring the Performance of Government Departments](https://www.nao.org.uk/reports/measuring-the-performance-of-government-departments/) — UK National Audit Office
- [Taking the measure of government performance](https://www.nao.org.uk/reports/taking-the-measure-of-government-performance/) — UK National Audit Office

## When a measure acquires power

Charles Goodhart and Donald Campbell offered related warnings from different intellectual traditions in the 1970s. Goodhart's work on monetary management examined the instability of relationships when policymakers tried to use statistical regularities as instruments of control. The relevant paper was presented in the mid-1970s and published in a Reserve Bank of Australia conference volume. Campbell's paper on assessing planned social change warned that quantitative social indicators could come under pressure when they became consequential tools of public decision-making.

Their arguments overlap, but they should not be treated as a single universal law. Goodhart was writing in a monetary-policy setting; Campbell was considering the use of indicators to assess social programmes. Neither work demonstrates that every target fails or that every performance measure must be manipulated. They identify a risk that deserves testing whenever an institution rewards, funds or judges behaviour through a proxy.

The mechanism is understandable without specialist mathematics. First, a public purpose is too complex or slow to observe directly. Second, the institution chooses an indicator that is easier to collect. Third, the indicator becomes a target linked to rewards, punishment, budgets or reputation. Fourth, people change their behaviour in response. Some changes genuinely advance the purpose. Others make the number look better by changing classification, timing or who gets served.

Suppose a fictional agency promises to answer ninety-five percent of complaints within a day. Staff might solve problems more quickly, which would be an improvement. Alternatively, the agency might send automatic acknowledgements and label them completed responses. Both strategies could improve the same reported statistic. The difference can be found only by examining what people received, not merely how the dashboard moved.

Sources:
- [Conference Volumes: Papers in Monetary Economics, Volume I and II (1976), including Goodhart's Problems of Monetary Management: The U.K. Experience](https://www.rba.gov.au/publications/rdp/1990/9013/conference-volumes.html) — Reserve Bank of Australia
- [Assessing the Impact of Planned Social Change — original 1976 paper reprinted with permission](https://jmde.journals.publicknowledgeproject.org/index.php/jmde_1/article/download/297/292/988) — Journal of MultiDisciplinary Evaluation / The Evaluation Center

## A child in school is not necessarily a child learning

Education provides a concrete illustration. Getting children into school matters; access to instruction can transform life opportunities. But enrolment alone does not tell us whether children can read, understand numbers or continue learning. A successful expansion of access and a failure to improve learning can coexist. It would be wrong to dismiss the first because the second remains unresolved.

The World Bank and UNESCO Institute for Statistics developed the Learning Poverty measure to bring schooling and learning into the same discussion. It combines information about children who are out of school with reading proficiency among those in school. The point is not to produce a magical single score. It is to make explicit that presence and learning are not identical dimensions of the educational experience.

Even this more thoughtful design has assumptions and limitations. For purposes of the measure, out-of-school children are counted as learning-deprived. Minimum proficiency marks a threshold but does not by itself describe how far below that threshold a child is. Data availability and comparability also affect what can be concluded. The methodology describes complementary measures of gap and severity precisely because a headcount does not convey every aspect of learning deprivation.

Consider what follows for a reader of an education report. Enrolment can answer an access question. Attendance can help answer whether children reach classrooms consistently. Reading proficiency can answer something about learning. Distributional breakdowns can reveal which populations remain excluded. None is automatically a substitute for the others. An institution improves the public conversation when it names which question each number answers and refuses to inflate the answer beyond its scope.

Sources:
- [What is Learning Poverty?](https://www.worldbank.org/en/brief/2021/04/28/what-is-learning-poverty) — World Bank / UNESCO Institute for Statistics

## The public service that cannot be priced

Many public institutions produce services without an obvious market price. A court is meant to deliver justice, not simply close files. A disaster agency is meant to improve preparedness and response, not merely issue plans. A hospital provides care whose value cannot be inferred from a count of procedures alone. These purposes are multidimensional and may unfold over different time periods.

The UK National Statistician's 2025 independent review of public-services productivity discusses this challenge directly. Non-market output lacks market prices that reveal the value of services to citizens, and quality adjustment is difficult when services produce several outcomes simultaneously. The review does not say that productivity measurement is impossible or that official statistics are worthless. It explains why methodological decisions, definitions and quality assumptions matter.

Imagine two fictional clinics each report one hundred consultations. The first might serve straightforward cases with excellent follow-up. The second might support people with severe or complicated needs. Equal throughput does not establish equal quality, effort or benefit. Likewise, fewer appointments may reflect reduced demand, longer consultations, constrained capacity or a shift in the case mix. The number can be a useful starting point while leaving the real explanation open.

A more sophisticated indicator may reduce some blind spots but introduce others. Weighting by case severity requires a model of severity; adjusting for quality requires assumptions about what quality means. A simple indicator may be transparent while incomplete. A composite score may appear comprehensive while concealing its own weighting decisions. In some settings a small group of clearly labelled measures is more honest than a single supposedly definitive ranking.

Sources:
- [National Statistician's Independent Review of the Measurement of Public Services Productivity](https://uksa.statisticsauthority.gov.uk/publications/national-statisticians-independent-review-of-the-measurement-of-public-services-productivity/pages/4/) — UK Statistics Authority

## The behavioural cost of an important target

Performance targets can focus attention on urgent problems. A waiting-time measure can reveal dangerous delays. A crime-recording measure can make a community's safety concerns visible. The risk emerges when the definition of success permits behaviour that optimises the metric without improving the service the metric is meant to represent.

The Northern Ireland Audit Office's 2018 guide to performance management describes this problem through examples of perverse incentives. It discusses the danger of patients being left in ambulances rather than admitted to emergency departments to avoid a waiting-time breach, and the possibility that pressure to reduce recorded crime may discourage appropriate recording. These are examples used in guidance, not evidence that all hospitals or police services behave that way or that the behaviour is widespread today.

The underlying mechanism lies in the boundary of the measure. If waiting time begins at a particular administrative moment, delaying that moment may make the official interval shorter without reducing how long the patient waits. If a crime target measures recorded offences, changing recording behaviour may alter the series even without a corresponding change in actual harms. A statistic can reflect changes in operations, rules, population and incentives at once.

This does not mean that improvements should be dismissed as gaming. Faster treatment may be real, and better reporting practices may produce apparent increases in crime even as safety improves. A careful analyst asks what else changed, how the data were collected and what independent evidence exists. The suspicion must be proportional to evidence; a metric should not be found guilty merely because it could in theory be manipulated.

Sources:
- [Performance management for outcomes: A good practice guide for public bodies](https://www.niauditoffice.gov.uk/files/niauditoffice/media-files/NIAO_performance%20management%20for%20outcomes.pdf) — Northern Ireland Audit Office

## The missing middle between announcement and outcome

A government can announce a reform, allocate money, create an agency, hire staff, issue a regulation and complete a training programme. Those are potentially meaningful steps. Yet they are not the same as making a public service more accessible, reliable or fair. The transformation from formal activity to experienced improvement runs through intermediate changes that are harder to capture.

The OECD's 2026 Government Unstuck report identifies a related measurement gap. Comparable governance data are relatively strong for inputs and processes, and a growing body of work tracks longer-term social impacts. More difficult to observe across countries are the intermediate outcomes connecting institutional reform to what citizens, businesses and public employees experience. The OECD proposes developing shared intermediate-outcome indicators rather than simply constructing another composite ranking.

This is an agenda for better evidence, not a finding that every government reform fails. A new permit system may initially show many applications processed, but the question readers care about could be whether decisions are timely, understandable and consistent. An emergency agency may meet its staffing target, but the key operational issue may be whether communities can access warnings and whether evacuation arrangements function under stress.

The middle also matters for interpreting apparent disappointment. Public experience may improve before national statistics register the change. Conversely, a national average may improve while particular regions or disadvantaged groups remain worse off. Differing survey and administrative records may both be informative because they measure different aspects of the system. Instead of discarding one, a serious review should ask what the disagreement reveals.

Sources:
- [Measuring progress towards Government Unstuck](https://www.oecd.org/en/publications/2026/10/government-unstuck_dcbcbf9a/full-report/measuring-progress-towards-government-unstuck_4a951766.html) — OECD

## A four-part diagnostic: Proxy, Incentive, Blind Spot, Verification

The following four questions are an editorial diagnostic for evaluating important statistics. They are not a validated empirical index or a claim of new peer-reviewed theory. Their purpose is to slow the leap from a changing number to a claim of completed success. They apply to public dashboards, social programmes, institutional performance reports and even corporate claims, provided the relevant context is supplied.

PROXY: What purpose is this number standing in for? Is it counting money allocated, staff employed, cases processed, service quality, citizen experience or a final outcome? An enrolment figure is an access measure; it is not a reading test. A count of hearings is procedural activity; it is not proof that participants influenced a decision. Ask what is directly measured before interpreting what may be indirectly suggested.

INCENTIVE: What behaviour does the target reward, and who controls the reporting definition? Could success be registered by moving the administrative start time, selecting easier cases or changing the unit counted? Conversely, does the target encourage a valuable improvement that would otherwise be neglected? Consider both possibilities. The fact that an incentive exists is not evidence of misconduct.

BLIND SPOT: Who and what disappear from the indicator? A national mean can conceal regional differences. A service completion rate can omit people who could not apply in the first place. A monthly number can overlook delayed harm. A count of outputs can obscure quality or long-term consequences. The blind spot is often found by asking whose experience would contradict the average.

VERIFICATION: What independent evidence would confirm that the underlying purpose improved? Look for another outcome measure, a data audit, a time series with stable definitions, disaggregated results or credible reports from service users. One additional measure will not always settle a dispute; sometimes multiple imperfect views are needed. But evidence obtained through a different method can reveal when the proxy and the goal begin to diverge.

The diagnostic is designed to be revisable. Some institutions will pass every question. Some indicators will fail because data are genuinely unavailable rather than because anyone intended to mislead. A useful framework recognises its own uncertainty: it opens an inquiry but does not produce a verdict without evidence.

Sources:
- [Performance management for outcomes: A good practice guide for public bodies](https://www.niauditoffice.gov.uk/files/niauditoffice/media-files/NIAO_performance%20management%20for%20outcomes.pdf) — Northern Ireland Audit Office
- [National Statistician's Independent Review of the Measurement of Public Services Productivity](https://uksa.statisticsauthority.gov.uk/publications/national-statisticians-independent-review-of-the-measurement-of-public-services-productivity/pages/4/) — UK Statistics Authority
- [Measuring progress towards Government Unstuck](https://www.oecd.org/en/publications/2026/10/government-unstuck_dcbcbf9a/full-report/measuring-progress-towards-government-unstuck_4a951766.html) — OECD

## The strongest case against distrusting targets

There is a powerful objection to any argument about measurement traps: without numbers, institutions can escape accountability through rhetoric. Professional judgement may be necessary, but it can also shelter complacency and favour those with influence. Anecdotes can be selected. Errors can be explained away. A transparent target can force a service to answer questions it would prefer to ignore.

The National Audit Office's work takes this positive role seriously. Well-designed performance measures can support public accountability, resource allocation and institutional learning. A target may direct attention to long-neglected delays or inequalities. In a system with little existing data, an imperfect indicator may be far better than a promise whose fulfilment cannot be checked.

This counterargument is correct in a way that must change the final claim. The answer is not to replace measurement with intuition or to demand impossible perfection. Nor is a complex model automatically superior to a simple, comprehensible number. A measure with clear definitions, independent validation and sensible incentives may deserve substantial confidence. Targets are not harmful merely because someone tries to meet them.

The more defensible position is narrower: measures need to remain accountable to purposes outside themselves. They can help institutions learn when they are treated as evidence rather than as symbolic achievements. Good measurement leaves room for an indicator to be criticised, refined or retired when it stops tracking the goal. Numbers can strengthen democracy and institutional competence precisely because people are allowed to question their meaning.

Sources:
- [Measuring the Performance of Government Departments](https://www.nao.org.uk/reports/measuring-the-performance-of-government-departments/) — UK National Audit Office
- [Taking the measure of government performance](https://www.nao.org.uk/reports/taking-the-measure-of-government-performance/) — UK National Audit Office

## Where this framework could fail

Some underlying purposes are closely aligned with measurable actions. If a service promises that a document will be issued within a statutory period, a verified count of on-time decisions may directly reflect an important part of the obligation. It may still miss accessibility or fairness, but that does not make the service-time measure invalid. The warning is weaker when the proxy closely matches the goal and the reporting process is independently audited.

Likewise, evidence can show that a target has improved alongside several independent outcomes, across groups and over time, without significant distortion of collection practices. In such a case the reported success should be recognised. An analytic tool that always discovers failure is not an effective diagnostic; it is a rhetorical device masquerading as one.

Sometimes there is no reliable direct measurement available at reasonable cost. A public-health authority may have to act quickly on provisional surveillance data. A disaster agency may rely on forecasts and preparedness indicators before a rare event allows direct observation of response performance. In these settings, acknowledging the proxy's limits while using it can be responsible practice. Waiting for perfect evidence may carry greater costs.

The practical boundary is simple: the framework can identify where interpretation requires caution, but it cannot itself reveal whether a programme succeeded. That verdict depends on case-specific evidence and on explicit decisions about the social outcomes being valued. The four questions must not be used to imply an original survey, an expert audit or a comparative causal finding that was never conducted.

Sources:
- [National Statistician's Independent Review of the Measurement of Public Services Productivity](https://uksa.statisticsauthority.gov.uk/publications/national-statisticians-independent-review-of-the-measurement-of-public-services-productivity/pages/4/) — UK Statistics Authority
- [Measuring progress towards Government Unstuck](https://www.oecd.org/en/publications/2026/10/government-unstuck_dcbcbf9a/full-report/measuring-progress-towards-government-unstuck_4a951766.html) — OECD

## What would someone actually notice?

The attraction of a number is that it appears to finish an argument. A target has been reached, and the line on the graph has moved in the desired direction. But the public purpose may still be partly outside the picture. The honest response is not automatic scepticism. It is curiosity about what the measure has captured and what would change if the underlying purpose had genuinely been served.

If a hospital improves its waiting-time statistic, would patients receive care sooner and more safely? If a school improves attendance, do more children learn? If a government establishes a new agency, do people and businesses experience a more reliable service? Each question needs its own evidence and time horizon. The answers need not fit into a single dashboard.

A number can be a remarkable instrument for accountability when it invites scrutiny rather than preventing it. Before accepting it as a verdict, ask what it measures, what incentives surround it, who is invisible and what independent evidence could test it. The final question is not whether the number looks better. It is whether the people for whom the number was created would recognise the improvement.

Sources:
- [Measuring progress towards Government Unstuck](https://www.oecd.org/en/publications/2026/10/government-unstuck_dcbcbf9a/full-report/measuring-progress-towards-government-unstuck_4a951766.html) — OECD
- [Taking the measure of government performance](https://www.nao.org.uk/reports/taking-the-measure-of-government-performance/) — UK National Audit Office

## About this article

This article reflects the public sources available on the publication and update dates shown above. For developing stories, follow the linked source pages for subsequent statements or decisions.

## Sources and further reading

- [Conference Volumes: Papers in Monetary Economics, Volume I and II (1976), including Goodhart's Problems of Monetary Management: The U.K. Experience](https://www.rba.gov.au/publications/rdp/1990/9013/conference-volumes.html) — Reserve Bank of Australia (published 1976) (accessed 2026-11-03)
- [Assessing the Impact of Planned Social Change — original 1976 paper reprinted with permission](https://jmde.journals.publicknowledgeproject.org/index.php/jmde_1/article/download/297/292/988) — Journal of MultiDisciplinary Evaluation / The Evaluation Center (published 2011-02) (accessed 2026-11-03)
- [What is Learning Poverty?](https://www.worldbank.org/en/brief/2021/04/28/what-is-learning-poverty) — World Bank / UNESCO Institute for Statistics (published 2021-04-28) (accessed 2026-11-03)
- [Performance management for outcomes: A good practice guide for public bodies](https://www.niauditoffice.gov.uk/files/niauditoffice/media-files/NIAO_performance%20management%20for%20outcomes.pdf) — Northern Ireland Audit Office (published 2018-06) (accessed 2026-11-03)
- [National Statistician's Independent Review of the Measurement of Public Services Productivity](https://uksa.statisticsauthority.gov.uk/publications/national-statisticians-independent-review-of-the-measurement-of-public-services-productivity/pages/4/) — UK Statistics Authority (published 2025-03) (accessed 2026-11-03)
- [Measuring progress towards Government Unstuck](https://www.oecd.org/en/publications/2026/10/government-unstuck_dcbcbf9a/full-report/measuring-progress-towards-government-unstuck_4a951766.html) — OECD (published 2026-10) (accessed 2026-11-03)
- [Measuring the Performance of Government Departments](https://www.nao.org.uk/reports/measuring-the-performance-of-government-departments/) — UK National Audit Office (published 2001-03-22) (accessed 2026-11-03)
- [Taking the measure of government performance](https://www.nao.org.uk/reports/taking-the-measure-of-government-performance/) — UK National Audit Office (published 2010-07-14) (accessed 2026-11-03)