The Fractured Lens

BelongingUnknowing

Engagement SurveysA Loyalty Test Dressed as Listening

In this essay 4 sections

On a scale from zero to ten, how likely are you to recommend this organisation as a place to work? The question arrives once or twice a year, usually by email, with a note saying the engagement survey takes ten minutes and that responses are anonymous. Almost everyone who has worked in an organisation of any size has answered it. Few know where it came from.

It came from customers. In December 2003 Fred Reichheld, a consultant at Bain & Company, published an article in the Harvard Business Review titled The One Number You Need to Grow. A single question about whether a customer would recommend a company, he argued, told you more about its prospects than a long satisfaction questionnaire. Customers who answered nine or ten were promoters, those who answered six or below were detractors, and the first percentage minus the second was the Net Promoter Score. Within a few years the question had moved through marketing departments, crossed the corridor into human resources, and settled at the top of the employee survey as the employee Net Promoter Score.

A customer who recommends a company brings it revenue. An employee asked the same question is being asked whether she will vouch for her employer in public, and that is a question about allegiance. The survey around it is presented as a way of listening to staff. Read closely, most of what it asks measures how committed people say they are to staying, and that shapes everything it is able to hear.

The Employee Net Promoter Score: A Customer Question at the Top of the Engagement Survey

The appeal of the score to the people who commission it is easy to understand. It is a single figure. It can be compared with last year’s and with a benchmark supplied by the vendor, broken down by division and team, put on a slide and reported to a board. It turns the diffuse and argumentative business of finding out what several thousand people think into a number that someone senior can own. When the number rises, someone takes credit. When it falls, someone is asked to produce a plan.

What the score cannot carry is reasons. A seven from someone who loves the work and despairs of the leadership looks identical to a seven from someone who is indifferent to both. The employee Net Promoter Score was designed by a consultant to predict whether customers would bring in more customers, and its critics in marketing have disputed for twenty years whether it does even that. In HR it was asked to do something else: to stand in for the condition of a workforce. In 2014 the employment relations scholar John Purcell reviewed the engagement field in an article titled Disengaging from Engagement. He found a concept with no settled definition, a body of research that cast the unengaged as a problem to be corrected, and an approach that left conflict at work out of the picture. Reducing engagement to a single score, he wrote, was especially worrying.

The single score has prospered anyway, because its virtues are virtues for the people reading it. Nobody filling in the survey has asked for their working life to be summarised as an integer between minus a hundred and a hundred. The integer serves the reader.

Its arithmetic has a bias built in. Only nines and tens count in favour; everything from zero to six counts against; sevens and eights count for nothing. A workforce in which every person answered eight, people who like their jobs and would say so to a friend, scores zero. The scale was built to separate enthusiasts from the rest, because for a consumer brand only enthusiasts bring in new customers. Applied to staff, it treats contentment as absence. The steady professional who does good work, has no complaints worth making and declines to perform excitement about her employer is, for the purposes of the number, nobody. The score registers allegiance only when it is loud.

Exit, Voice and Loyalty: What Engagement Surveys Measure

In 1970 the economist Albert Hirschman published a short book with a long reach, Exit, Voice, and Loyalty. When an organisation begins to decline, he argued, its members have two ways to respond. They can leave, which he called exit. Or they can stay and complain, argue, petition and try to change things from inside, which he called voice. Loyalty is what tips a person towards the second. A loyal member stays longer than her dissatisfaction alone would justify, and spends that time objecting, and the organisation gains because voice tells it what is wrong while there is still time to act.

Set the engagement survey against Hirschman and it falls across his categories at an odd angle. It is offered as a channel for voice. The questions that carry the most weight, whether you would recommend the place and whether you expect to be here in two years, are readings of loyalty. And loyalty, in Hirschman’s account, has its value because of what it produces: the irritating persistence of someone who cares enough to argue. A survey that records loyalty as a score has taken the source of voice and filed it as a result. The organisation learns how loyal its people say they are. What their loyalty would have prompted them to say goes unrecorded.

There is a free-text box at the end for that. In most large organisations its contents are sorted by software into themes and sentiment before anyone senior reads them, which converts them back into something closer to a score. Voice is expensive to hear. It arrives in sentences, it names people and decisions, it comes at inconvenient times and it asks for things. A theme labelled “communication” asks for nothing.

Hirschman’s framework has a second edge, and it cuts deeper. If voice has been absorbed into a loyalty reading, exit becomes the only signal the organisation cannot misread. The person who answers eight in March and resigns in June has given two answers. One of them was shaped by everything a survey is shaped by: who might see it, what it might cost, what the manager needs it to say. The other was given with her feet, and there is no benchmark for it. Exit is the survey nobody can fill in on someone else’s behalf. An organisation that treats its engagement score as the truth about its people, and its attrition figures as a separate matter for recruitment, has arranged to receive the true answer only after the person giving it has gone.

Exit is also not always open. When jobs are hard to find, Hirschman’s balance shifts: dissatisfied people stay because they must, and their discontent goes either into voice or into silence. Gallup reported that optimism about the job market in Australia and New Zealand fell twelve points during 2025. In such conditions a stable or rising engagement score may record nothing more than a workforce with fewer places to go, answering the loyalty question the way people answer when leaving is not a realistic option. The number cannot distinguish commitment from captivity. Neither, from the outside, can anyone else.

Engagement Survey Anonymity and the Honest Answer

The word engagement had an earlier life, and the part of it worth keeping concerns safety. In 1990 William Kahn studied summer camp counsellors and the staff of an architecture firm, and described engagement as something people did moment by moment: bringing more of themselves into their work or holding more back, depending on what the situation offered. One of the conditions he identified was psychological safety, whether a person could show herself without fear of damage to her standing or career.

An engagement survey cannot measure that condition, because its own accuracy depends on it. Where people feel safe the survey will be broadly honest. Where they do not, it will report what they judge it safe to report, and nothing in the numbers will reveal which kind of place produced them. The instrument assumes the condition it ought to be testing. Kahn’s third condition, availability, whether a person has anything left to bring, fares no better: the survey lands in the same inbox as everything else, often in the weeks after a restructure, and is answered by people deciding how little of themselves they can get away with giving it.

People completing the forms understand this better than the people commissioning them. In October 2024 the analytics firm Visier asked a thousand American employees about the engagement surveys they took. Close to half said they felt pressure to hold back at least occasionally, and more than a third did not believe the surveys were ever anonymous. The subjects they found hardest to be honest about were the ones the survey most wanted to understand: satisfaction with the job, the performance of leadership, the relationship with their own manager. Their leading reason was distrust of how the data would be handled. The next was a belief that nothing would change.

That second reason has been taught. The survey cycle usually ends in an action plan, and the action plan is built from what can be changed without changing anything that matters to the people who commissioned it: a new recognition scheme, say, or more frequent updates from leadership. The posters that follow announce “you said, we did”. People notice the distance between what they said and what was done, and they adjust. Some stop writing in the free-text box. Others write in it with a care that makes it useless, phrasing every concern so that it could not be traced or resented. The organisation reads the declining volume of comment as a sign that things have improved.

The distrust has grounds. In a team of six, a response broken down by length of service and location may identify its author. Survey vendors set minimum group sizes to prevent this, and the people answering rarely know what the minimum is, whether it was respected, or who holds the underlying file. Then the results travel. They are cut by team and sent to managers, who are asked to discuss them and produce action plans, and whose standing comes to depend on the number. Reichheld himself has said that people have been disciplined over scores they could not control, and has argued for separating the score from pay. A team that watches its manager being held to its score learns what the score is for, and begins to answer on her behalf. That is the most humane kind of dishonesty available, and the survey has no way to detect it. Organisational Psychology traces how measurement of this kind changes what it measures; the engagement survey is the case in which the people being measured can watch it happen, and adjust.

Engagement as Belonging: The Survey as a Membership Ritual

Answering nine or ten is a statement of membership. Answering six places you, in the survey’s own vocabulary, among the detractors. The word was coined for customers who might damage a brand by speaking badly of it. Carried into the workplace it names something closer to an enemy within, a person whose presence lowers the score that someone senior has to explain.

Few people think of their answers in these terms, and they do not need to, because the categories do the thinking. Promoters are believers and detractors are a risk; passives barely count. When a team’s score falls, attention follows, usually a listening session or a visit from someone in HR with a slide deck of themes. The attention is offered as care, and often meant as care. It also tells the team that the number was noticed and that someone will be asking about it. What a person learns after a few years of this is which answer ends the conversation.

Over time the survey settles into a ritual of renewal, a periodic occasion on which members reaffirm their belonging in the language of measurement. Its function becomes less to find out what people think than to give them a regular, low-cost way of saying that they are still in. The illusion of psychological safety works the same way: a condition that can only be lived is turned into a programme that can be audited, and belonging becomes something performed for the people doing the audit. The engagement survey is one of the places where that performance is scheduled.

Gallup’s State of the Global Workplace reports that one employee in five was engaged in 2025, the lowest share since 2020, and that engagement among managers fell further and faster than among the people they manage. The global figure has never reached a quarter in all the years Gallup has published it. For an instrument meant to improve the condition it measures, two decades of that would count as failure. The surveys continue, are renewed and are extended into monthly pulses, which suggests the organisations buying them are getting something the global figure does not record.

The extension has a cost the buyers rarely count. Reichheld, watching his question spread to every receipt and delivery confirmation, came to call the flood of rating requests a tragedy of the commons: each organisation’s survey is reasonable on its own, and together they teach people to stop answering carefully. Inside a workplace the commons is smaller and the effect is faster. A monthly pulse asks people to report on their state more often than their state changes, and the reports become reflexes. The numbers keep arriving, with less and less behind them, and their regularity is mistaken for reliability.

Part of what they get is reassurance: evidence that leadership asked, a number for the board, and a ritual that binds people a little more firmly each time they perform it. The other part is a familiar kind of unknowing, the comfort of a measure that makes it unnecessary to go and find out. Instruments built to surface contradiction, by asking people to hold two statements side by side and noticing where they agree with both, make that comfort harder to keep, which is why the AI Transition Diagnostic was built around pairs instead of a single score. A pair cannot be averaged into allegiance.

None of this settles what an organisation of thousands should do in place of the survey, and it may be that the survey survives because nobody has found a way of hearing voice at that scale. The question at the top of the form is still the one Reichheld wrote for customers. The people answering it are still being asked whether they will vouch for the place, by the place itself, on a form it designed, and most of them are still saying something close to yes. What they would say to a question that could not be traced or scored is not something any engagement survey has been designed to find out, and it is not obvious that the organisations running them would want it to.

New essays by email

The next essay when it is published. Nothing else.