Measuring satisfaction with public services

A citizen renews a trade licence online in four minutes flat, gets a pop-up asking her to rate the experience, and taps five stars without thinking twice. National satisfaction figures nudge upward for the third year running. Meanwhile, the queue outside the licensing office two streets over is still three hours long, and the people in it were never asked anything. This is the quiet fraud at the heart of most public-sector satisfaction measurement: it counts the voices that were easy to hear, not the experience that was actually delivered.

Measuring satisfaction with public services fails when governments borrow tools built for markets where customers can walk away. Citizens usually can't switch providers, so a satisfaction score doesn't measure delight — it measures resignation. The fix is to stop asking "how happy are you?" and start measuring three harder things: how much effort the service demanded, how fairly the citizen felt treated, and whether trust survived the interaction. Get those three right, and the annual survey becomes almost beside the point.

Why doesn't standard customer satisfaction measurement work for government?

Most satisfaction metrics — Net Promoter Score chief among them — were built on a simple assumption: the customer has somewhere else to go. If the coffee is bad, they walk next door. That threat of exit is what makes a satisfaction score meaningful. It reflects a real choice, made freely.

The economist Albert Hirschman named this dynamic in his 1970 book Exit, Voice, and Loyalty: Responses to Decline in Firms, Organizations, and States (Harvard University Press), arguing that when exit is unavailable, people fall back on voice — complaining, escalating, organising — or on quiet resignation. Government sits squarely on the wrong side of that equation. A resident can't switch national ID authorities, choose a rival passport office, or take her tax return to a competitor. Exit is off the table, so what a satisfaction survey usually captures instead is the residue of people who've stopped expecting better, sitting alongside a smaller group of people angry enough to still complain. That's not a satisfaction score. It's a resignation index wearing a satisfaction score's clothes.

This is why public bodies routinely report high satisfaction and low trust in the same year. The two numbers aren't contradictory — they're measuring different things, and only one of them tells you anything useful about whether the service is actually working. For a longer view of how citizens weigh procedural experience against outcomes, see our companion piece on measuring satisfaction with public services.

What should governments measure instead of a happiness score?

Governments get a far more honest picture by measuring effort, fairness, and trust separately rather than folding them into a single satisfaction number. Each behaves differently, breaks for different reasons, and needs a different fix.

Effort is the easiest to act on and the most consistently predictive. In their 2010 Harvard Business Review article Stop Trying to Delight Your Customers, Matthew Dixon, Karen Freeman and Nicholas Toman — then researching for the Corporate Executive Board — found that reducing the effort a customer has to expend does more for loyalty than trying to exceed their expectations. In a public-service context, "loyalty" isn't the right word, but the mechanism holds: a citizen who had to visit three counters, upload the same document twice, and call a hotline that didn't answer will carry that friction into every future interaction with the state, regardless of how the final outcome felt.

Fairness matters even more, and it's measured almost nowhere. The psychologist Tom R. Tyler's research on procedural justice, laid out in his 1990 book Why People Obey the Law (Yale University Press), found that people's willingness to accept a decision — a fine, a rejected application, a denied benefit — depends far more on whether the process felt fair and respectful than on whether the outcome favoured them. A citizen denied a subsidy by a transparent, well-explained process will often accept it. A citizen granted the same subsidy through an opaque, rude, or arbitrary-feeling process will remember the arbitrariness long after the money has landed. Public bodies that measure only outcome satisfaction miss this entirely.

Trust is the lagging indicator that ties the other two together, and it's worth tracking over years, not survey cycles. Pew Research Center's long-running series on public confidence, most recently updated in its Public Trust in Government: 1958–2024 report, has tracked American trust in federal government continuously since 1958 — and shows how slowly trust rebuilds compared with how quickly it erodes. That asymmetry is not an accident of politics. It's behavioural, and it shapes how any measurement system should be designed.

How does the peak-end rule shape what citizens remember about a service?

People don't average their experience of a service — they remember the most intense moment and the ending, and judge the whole interaction by those two points. This is the peak-end rule, first demonstrated by Daniel Kahneman and Donald Redelmeier in their 1996 study of colonoscopy patients, published in the journal Pain, which found that patients' retrospective ratings of a painful procedure depended far more on the pain at its worst moment and its final minutes than on its total duration or cumulative discomfort.

Apply that to a citizen renewing a residency permit. She might spend forty pleasant minutes navigating a well-designed portal, then hit a confusing payment error at the very last step that takes three days and two phone calls to resolve. Average the whole journey and it looks mostly fine. But she won't remember the forty smooth minutes. She'll remember being stuck at the finish line, and that memory — not the average — is what she reports on next year's survey, and what she tells her neighbours.

This has a direct implication for where governments should concentrate design effort: the final step of any public-service journey — confirmation, payment, handoff to a human, the moment a decision is delivered — carries disproportionate weight in the citizen's memory. Mapping the journey stage by stage, rather than department by department, is the only way to find where those peaks and endings actually sit. That's the logic behind structured CX journey mapping: it forces the organisation to see the sequence a citizen actually experiences, not the sequence of internal handoffs that produced it.

Why does loss aversion make public trust so hard to rebuild?

A single bad experience with a government service erases the goodwill built by many good ones, because losses are felt roughly twice as intensely as equivalent gains. This is loss aversion, first formalised by Daniel Kahneman and Amos Tversky in their 1979 prospect theory paper published in Econometrica, and it explains a pattern every public-sector CX practitioner recognises: years of steadily improving digital services can be undone almost overnight by one high-profile failure — a benefits system that wrongly cuts off payments, a data breach, a queue that goes viral on social media.

For measurement, the practical consequence is this: an average satisfaction score can mask a small number of severe negative outliers that matter far more to institutional trust than the average suggests. A citizen whose pension payment was wrongly withheld for six weeks doesn't care that the average processing time across the population improved. She experienced the loss, and loss aversion means that experience will outweigh a dozen smooth renewals in how she talks about the institution afterwards. Measurement systems built only around means and medians will always underweight these events. They need a separate tier that flags and escalates severe negative experiences the moment they occur, not months later in an annual report.

How should a public body actually build a measurement system that works?

Measurement only earns its keep if it changes what the organisation does next — the sequence below moves a public body from a once-a-year survey to a system that catches problems while they're still fixable.

1. Map the journey, not the department. Most public-sector satisfaction data is organised around the unit that ran the survey — the licensing department, the tax authority, the health directorate. Citizens don't experience government that way; they experience a sequence of stages that often cuts across three or four agencies. Build the measurement architecture around that sequence first.
2. Capture feedback at the moment of the interaction, not months later. An annual citizen satisfaction survey tells you what people remember, filtered through the peak-end rule and whatever happened in the news that week. In-the-moment feedback — a short prompt immediately after a service is delivered — tells you what actually happened.
3. Separate effort, fairness, and outcome into distinct questions. A single overall satisfaction score collapses three different signals into one number that nobody can act on. Ask "how much effort did this take you?" as a distinct question from "did the outcome match what you needed?" and "were you treated fairly and given a clear explanation?"
4. Segment by vulnerability, not just by channel. An elderly citizen without reliable internet access and a small-business owner filing digitally are having entirely different experiences of the "same" service. Aggregate satisfaction figures routinely hide the fact that the people struggling most are also the least likely to respond to a survey at all.
5. Close the loop publicly. Nothing degrades trust in feedback mechanisms faster than a citizen who takes the time to flag a problem and never hears what happened to it. Publishing what changed as a result of citizen feedback — even small, unglamorous fixes — does more for perceived legitimacy than the fix itself. Our piece on closing the voice-of-customer feedback loop sets out how to build that discipline without it collapsing into box-ticking.
6. Govern it properly. A measurement system without clear ownership, escalation paths, and accountability for acting on findings will drift into becoming a compliance exercise within eighteen months. A defined CX governance structure is what keeps the data connected to decisions rather than filed in a quarterly report nobody reads.

What goes wrong when public bodies measure satisfaction badly?

The failure modes are consistent enough across markets and sectors that they're worth naming explicitly, because most are avoidable with basic design discipline rather than more budget.

• Survivorship bias in the sample. Digital surveys reach people who successfully completed the digital service. The citizen who abandoned the process in frustration, or who never had the access to attempt it, is invisible in the data — and is usually the most dissatisfied person in the population.
• A single annual measurement point. One survey a year cannot detect a service that degrades in March and recovers by November. By the time the annual result is published, the problem it should have flagged is either long resolved or has metastasised.
• Targets that reward the score instead of the service. When frontline teams are measured on satisfaction ratings rather than on resolving the underlying issue, staff learn — consciously or not — to manage the survey rather than the outcome, asking for high ratings before a case is genuinely closed.
• No route from insight to action. Feedback collected by a central unit with no authority over the frontline teams that caused the friction dies in a dashboard. The people who can fix the queue, the form, or the script never see the data that describes the problem they're causing.
• Treating complaints as noise rather than signal. A spike in complaints about a specific step is one of the richest, cheapest, and most immediate satisfaction signals a public body has. Many treat it purely as a customer-service workload problem instead of the leading indicator it actually is.

Fixing these failure modes doesn't require a bigger survey. It requires treating feedback management as operational infrastructure rather than an annual reporting obligation, and building the kind of end-to-end citizen experience discipline that connects what's measured to what gets changed. Public bodies that want a structured starting point can benchmark their current measurement maturity against a recognised framework using a tool such as the CX Maturity Assessment, which scores an organisation across the building blocks — governance, measurement, and journey design among them — that separate a functioning feedback system from a decorative one. For a broader view of how public institutions are adapting service delivery under digital transformation pressure, see our overview of public-sector customer experience and digital transformation.

What does good look like, in practice?

The public bodies that get this right share a habit more than a technology: they treat every service interaction as a data point about the state's credibility, not a customer-service metric to be optimised in isolation. Estonia's digital-government infrastructure is often cited for its technical elegance, but the more instructive part is procedural — citizens can see who accessed their data and why, which converts an opaque bureaucratic process into a transparent, accountable one. That transparency is itself a fairness intervention, not just a technology feature, and it's the kind of design choice that shows up in trust scores long before it shows up in satisfaction surveys.

The lesson generalises well beyond digital government. A service doesn't need to be fast to be trusted. It needs to be legible — the citizen needs to understand what's happening, why, and what happens next. Effort measurement tells you where the process is heavy. Fairness measurement tells you where the process feels arbitrary. Trust measurement tells you whether the sum of those experiences is building or eroding confidence in the institution over time. None of the three substitutes for the others, and none of them is captured by a single ten-point satisfaction question asked once a year.

The state doesn't need citizens to love it. It needs them to be able to predict how they'll be treated — and to be right. A measurement system that tracks effort, fairness, and trust as distinct signals, captured close to the moment of interaction and connected to real authority to act, will tell a public body far more about its own legitimacy than any satisfaction score ever has. Build that system, and the survey stops being the report card. It becomes the early-warning radar it was always supposed to be.