What Gets Measured Gets Managed and Everything Else Gets Lost

There is a seductive logic to measurement. If something can be tracked, surely tracking it reveals something true. If a number goes up, something has improved. If a number goes down, something needs fixing. We live, increasingly, inside this logic, not just in our organisations, where dashboards and performance frameworks multiply like Japanese knotweed, but in our personal lives, where wearables count our steps, score our sleep, and assign numerical values to our stress. We have not simply gained access to more information about ourselves. We have adopted a particular epistemology: that what is measurable is real, and what is real ought to be measurable.

This piece is not a polemic against data. Data is useful. The problem is a category error we have been making quietly, at scale, for some time. It’s the confusion of legibility with understanding. To make something legible is to render it visible and countable on your terms, not the thing’s own terms. Understanding requires context, history, judgement, and crucially, a willingness to sit with what cannot be captured in a figure. An organisation that mistakes the former for the latter will make decisions that look rational on a spreadsheet and feel catastrophically wrong on the ground. A leader who mistakes their dashboard for self-knowledge is not better informed. They are differently deceived.

The illusion of objectivity

Every act of measurement is an act of interpretation in disguise. The choice of what to measure, how to measure it, on what timescale, against what baseline, and with what instrument - each of these is a judgement call, saturated with assumptions. Data does not arrive scrubbed clean of human bias; it arrives already shaped by the questions we decided to ask and the things we decided not to count.

The political scientist James Scott explored this dynamic in his 1998 work Seeing Like a State. Scott observed that states and large organisations routinely simplify complex social realities in order to make them legible: taxable, administrable, plannable. Forest management reduced the diversity of an ancient ecosystem to a calculation of timber yield per acre. Urban planning replaced the organic complexity of working neighbourhoods with rational grids. In both cases, Scott argued, the process of making something legible destroyed the very features that made it function. The simplified, measurable version of the thing was not a better representation of reality. It was a different thing entirely, optimised for visibility rather than vitality.

"No administrative system is capable of representing any existing social community except through a heroic and greatly schematized [sic] process of abstraction and simplification. It is not simply a question of capacity, although, like a forest, a human community is surely far too complicated and variable to easily yield its secrets to bureaucratic formulae." James C. Scott

The same logic operates in schools, hospitals, universities, and corporations across the world. We generate data on the things we can measure - attendance, attainment scores, staff turnover rates, patient throughput, publication outputs - and we treat this data as though it were a faithful portrait of organisational health. It is not. It is a portrait of organisational legibility. The distinction matters enormously because the actions that improve legibility are not always the same as the actions that improve the underlying reality. Sometimes they are directly opposed.

Goodhart’s Law and the collapse of the measure

In 1975, the economist Charles Goodhart observed something that has since been compressed into a formulation so widely cited as to have almost lost its bite: when a measure becomes a target, it ceases to be a good measure. Goodhart was writing about monetary policy, but the principle generalises with uncomfortable precision to almost every domain in which measurement is used to drive human behaviour.

"When a measure becomes a target, it ceases to be a good measure." Charles Goodhart

The mechanism is not complicated. Identify a metric. Attach consequences - rewards, punishments, rankings, inspections - to that metric. The people being measured will, rationally, reorganise their efforts to optimise for the metric. The metric then no longer measures what it was originally designed to measure, because the behaviour it was designed to track has been replaced by behaviour designed to produce the metric. The indicator has been gamed, not necessarily dishonestly, and often not even consciously, but gamed nonetheless.

Donald Campbell, an American social psychologist, arrived at an equivalent formulation independently in 1979, arguing that,

“The more any quantitative social indicator is used for social decision-making, the more subject it will be to corruption pressures and the more apt it will be to distort and corrupt the social processes it is intended to monitor.” Donald Campbell

Campbell’s Law and Goodhart’s Law are, in essence, the same observation made from different disciplines: measurement, when consequential, controls what it is meant to observe.

The British education system offers examples so abundant they have ceased to shock. When league table position became the metric by which schools were judged, schools managed their league table position -  through decisions about which pupils to recruit, which subjects to offer, how to allocate teaching time, and how to interpret borderline grade decisions. The metric did not lie, exactly. It simply ceased to measure what a good school actually is. NHS waiting times present the same dynamic. When the four-hour accident and emergency target became the measure of emergency care quality, trusts found ways to meet the four-hour target that had no necessary connection to the quality of emergency care. Ambulances were held outside. Patients were treated in corridors that were technically not the emergency department. The number improved. The care did not.

None of this required bad faith. It required only that rational people respond rationally to the incentives they had been given. This is Goodhart’s real lesson: the problem is not corruption, it is design. If you build a system around a measurable proxy for what you actually care about, you should expect the proxy to be optimised rather than the thing itself.

The quantified self goes personal

The corporate world’s addiction to measurement has a personal counterpart that has crept up on us with the help of technology small enough to wear on our wrists. The quantified self movement, which is the systematic tracking of personal behaviour through wearable devices, apps, and self-monitoring tools promises the same thing that organisational dashboards promise: that more data produces better understanding, and better understanding produces better outcomes.

The philosopher Charles Taylor spent much of his intellectual life examining what authenticity actually means, and his conclusions are not comfortable for the data-tracking industry. In The Ethics of Authenticity, published in 1991, Taylor distinguished between two very different senses of knowing oneself. The first is what he called the horizon of significance, which he called the deep, often inarticulate background of values, commitments, and orientations that constitute who someone actually is. The second is the surface performance of selfhood: the curated, visible, measurable version of the self that can be presented and evaluated. Taylor’s worry was that modern culture increasingly rewards the second at the expense of the first, and in doing so produces selves that are highly optimised for external evaluation but progressively estranged from genuine self-understanding.

The quantified self is the technological apotheosis (great word by the way that I learned recently!) of exactly this tendency. A wearable device can tell you that you slept for six hours and forty-two minutes, that your resting heart rate was elevated, that you took 4,300 steps fewer than yesterday, and that your stress score (whatever the heck that means) was high. What it cannot tell you is whether the conversation you had at lunch was the most meaningful you have had in months, or whether the walk you didn’t take would have freed a thought that has been stuck for weeks, or whether the fatigue you are feeling is the productive exhaustion of genuine creative effort or the hollow depletion of work without purpose. The device measures what it can detect. You experience what actually happens. These are not the same thing, and no quantity of data bridges the gap. 

The deeper problem is that measurement changes behaviour, not always for the better, and not always consciously. Once you are tracking your sleep score, the quality of your sleep becomes entangled with the quality of your score. Once you are counting steps, movement becomes instrumentalised. The experience of walking, which has served as a site of thought, encounter, and pleasure for as long as humans have walked is partially colonised by its function as a data point. You are not walking. You are generating steps. The activity has not changed, but your relationship to it has, and that relationship is part of what the activity is.

Frederick Winslow Taylor’s ghost

The quantified self is not a new idea dressed in new technology. It is a very old idea, one that was fought over vigorously in the early twentieth century and which we seem to have forgotten we resolved.

Frederick Winslow Taylor was an American engineer who, in the late nineteenth and early twentieth centuries, developed what he called scientific management: the systematic analysis and optimisation of industrial work through measurement, standardisation, and time-and-motion study. Taylor believed that there was, for every task, a single best method, and that this method could be identified through careful observation and measurement. Workers who deviated from it were inefficient; workers who followed it were productive. Management’s job was to discover the optimal method and enforce it.

“What we are all looking for, however, is the readymade, competent man; the man whom some one else has trained. It is only when we fully realize [sic] that our duty, as well as our opportunity, lies in systematically cooperating to train and to make this competent man, instead of in hunting for a man whom some one else has trained, that we shall be on the road to national efficiency.” Frederick Winslow Taylor

Taylorism was fiercely contested by workers and trade unions who understood, even without the philosophical vocabulary to fully articulate it, that what was being lost in the process of optimisation was the judgement, craft knowledge, and human intelligence that skilled work actually depends on. Taylor saw these as inefficiencies to be eliminated. His critics saw them as the thing itself.

The philosopher Matthew Crawford, writing in The World Beyond Your Head in 2015, has argued that embodied, skilled, practical knowledge - the kind that a craftsperson, a surgeon, or an experienced teacher develops over years - is not a primitive precursor to abstract, quantifiable expertise. It is a distinct and irreplaceable form of intelligence, one that cannot be decomposed into measurable components without being destroyed. Crawford’s argument is that the denigration of this kind of knowing, in favour of abstract, transferable, legible knowledge, impoverishes both individuals and institutions.

I saw this firsthand in Murano the other week whilst on holiday. Murano, a small island across the water from

Venice, is world famous for its glass. Glassblowing masters often join the trade at 15 and train for 15-20 YEARS before becoming a ‘master’. Many don’t make it and become assistants. We watched Marco do a demonstration from the molten glass into a wonderful vase and small horse figurine. In just ten minutes, we watched an absolute craftsman in the most nonchalant manner possible, do something that AI or any data mechanism could never replicate. 

Taylorism was largely discredited as an approach to managing knowledge workers and creative labour. What is striking is that its fundamental logic -  break the activity into measurable components, identify the optimal parameters, track compliance, reward performance on the metric - has been completely rehabilitated in the form of data-driven leadership, personal productivity systems, and the quantified self. We resisted it when managers applied it to workers. We have welcomed it enthusiastically when we apply it to ourselves. The ideology has not changed. Only the direction of the stopwatch (or smartwatch) has.

What measurement cannot touch

There are aspects of human excellence that are not merely difficult to measure, they are structurally resistant to measurement, in the sense that the attempt to measure them changes or destroys what made them valuable. This is not a counsel of despair or mysticism. It is a precise philosophical claim with practical consequences.

Genuine moral attention - the capacity to see another person, a situation, or a problem clearly and without distortion - requires a quality of self-forgetfulness that cannot be achieved through deliberate effort or systematic method. Iris Murdoch called this quality of attention “love” in a technical sense: not sentiment, but the disciplined relinquishing of the self-centred perspective in favour of reality as it actually is. Her point was that ethical perception is not a skill that can be broken down into competencies and assessed against a rubric. It is a form of being, cultivated through practice, that either characterises how you engage with the world or it does not.

If we apply this to teaching or leadership, the implications are uncomfortable. The qualities that make a genuinely great teacher - the ability to notice when a particular student is struggling with a concept in a particular way, to adjust the register of an explanation mid-sentence, to create in a room the conditions under which a student who has never thought of themselves as capable decides, quietly, to try, or even that capacity to make abstract concepts concrete with personalised examples - none of this shows up in data. Lesson observation grades capture a proxy. Pupil progress data captures a downstream effect. The thing itself, what Aristotle called phronesis, practical wisdom, the capacity to perceive what a situation requires and respond with appropriate judgement is developed through years of experience, reflection, and human encounter. It cannot be dashboarded, and the attempt to reduce it to observable, measurable competencies produces something that looks like professional practice but progressively drives out the thing that made it alive.

The same is true of leadership. The decisions that matter most like how to navigate a genuine values conflict, how to hold a community together through serious difficulty, when to hold firm and when to change depend on a quality of judgement that data can inform but never replace. Leaders who have been trained to treat data as the primary input to decision-making are not better equipped for these moments. They are less equipped, because they have spent less time developing the forms of attentiveness, relational intelligence, and moral seriousness that these moments require.

Reclaiming judgement

None of this is an argument for ignorance. Data, used proportionately and with epistemological modesty, is genuinely useful. Knowing that your Year 9 cohort’s reading ages have not progressed in twelve months is important information. Knowing that your staff turnover rate has tripled in two years tells you something significant. The point is not to discard the information but to refuse it the authority it does not deserve - the authority to substitute for judgement rather than inform it.

The philosopher Onora O’Neill delivered the BBC Reith Lectures in 2002 under the title A Question of Trust. Her argument was precise and given that the lectures are now more than two decades old prescient to a degree that should embarrass us. O’Neill observed that the explosion of transparency and accountability mechanisms in British public life had not produced more trust. It had produced more compliance, more audit, more documentation, and more of what she called “the destruction of professional judgment.” The mechanisms designed to make institutions trustworthy had, perversely, colonised the time and attention of the professionals within them, leaving less space for the actual work on which trust depends.

“We have constructed an accountability system that makes things harder to trust, not easier.” Onora O’Neill

The solution she proposed was not the abolition of accountability but the careful reconstruction of what accountability is actually for. It’s not for the production of legibility for external observers, but the genuine answerability of professionals to the people they serve. The difference between these two things is the difference between a teacher who can produce a portfolio demonstrating their compliance with eight teaching standards and a teacher who can explain, in honest and specific terms, how they think about the children in front of them and what they are trying to do.

This is the distinction that the quantified self cannot navigate, because it is oriented entirely towards legibility. A sleep score makes your sleep visible to an app. It does not make you better rested, and it does not tell you what rest is for. A leadership dashboard makes your organisation’s performance visible to governors, trustees, or inspectors. It does not make your organisation better led, and it does not tell you what leadership is for.

What it means to reclaim judgement is not dramatic. It is the steady work of building the conditions in ourselves, in our teams, in our institutions under which the qualitative dimensions of excellent practice are taken seriously. It means treating the wisdom of experienced practitioners as a form of knowledge, not just a collection of opinions awaiting validation by data. It means creating time for the kinds of conversation - slow, exploratory, uncertain - in which genuine understanding develops. It means refusing to evaluate everything, because evaluation is not neutral: it changes what it touches, and not always in ways that serve the underlying purpose.

Aristotle, for all that he has been cited in every context imaginable, identified something in the concept of phronesis that has not become less true with age (I have written about this SOOO many times!). Practical wisdom is not the application of a rule or the execution of an algorithm. It is the capacity to perceive, in the specific texture of a real situation, what this moment requires and to act accordingly. It takes a long time to develop. It cannot be accelerated by a dashboard. And it remains, despite everything we have built to replace it, the thing that leadership and teaching and most of what matters in professional life actually depends on.

The data will tell you what it can see. The question is whether you are still capable of seeing what it cannot.

Key Takeaways

1. Goodhart’s Law is not an edge case, it is the default condition of any measurement system applied to human behaviour. The moment a metric becomes a target, the system begins to reorganise itself around producing the metric rather than achieving the underlying goal. This isn’t cynicism or corner-cutting; it is a predictable structural response. If you design a system that rewards measurable outputs, don’t be surprised when people optimise for measurable outputs.

2. Data gives you legibility, not understanding and these are not the same thing. To make something legible is to render it visible to an external observer on their terms, not yours. Understanding requires context, history, and judgement that no dataset carries with it. An organisation that confuses the two will make decisions that look rational on a spreadsheet and feel catastrophically wrong on the ground.

3. The quantified self is not a more honest self; it is a more visible one. Tracking your sleep, steps, and stress scores doesn’t reveal who you are; it reveals what your devices can detect. The aspects of human experience that matter most - moral courage, relational attentiveness, creative capacity - leave little to no data trail. Mistaking visibility for self-knowledge is one of the quieter confusions of our time.

4. Taylorism didn’t disappear but it went personal and called itself a wellness app. Frederick Winslow Taylor believed that human work could be broken into measurable units and optimised accordingly. That logic was contested and resisted when applied by managers to factory workers; it is now welcomed enthusiastically when we apply it to ourselves. The ideology hasn’t changed but what we are measuring has. 

5. Practical wisdom cannot be dashboarded. Aristotle’s phronesis is built through experience, reflection, and human encounter, not through performance metrics. The leaders and teachers who genuinely change lives rarely show up as outliers in the data. That should trouble us far more than it does.

6. Accountability systems that replace trust don’t build it, they hollow it out. Onora O’Neill argued that the proliferation of transparency and accountability mechanisms in public life has not produced more trust - it has produced more compliance theatre. When professionals spend increasing time demonstrating that they are doing their jobs well, they have less time to actually do them. At some point, the measurement becomes the work.

The data will always tell you what it can see. The harder question and the one that no instrument can answer for you is whether the things you most need to understand are the things your instruments are designed to detect. In most of what matters in leadership, teaching, and professional life, they are not. That is not a technical problem awaiting a better sensor. It is a philosophical problem awaiting a different kind of attention.