At the monthly cabinet meeting, the citizen services department presents a slide with one big number: 4.6 out of 5 stars. The mayor asks what anyone would ask: 4.6 of what?
Nobody in the room can answer precisely. The survey goes out through a generic link at the end of some conversations. Nobody knows how many people received it, how many answered or which report each response referred to. Nobody knows whether the person who gave five stars was rating the repair of their leak or the kindness of whoever took down their details. And there is no way to know whether the 300 people whose pothole was never filled are in that average or simply did not answer.
The 4.6 is a rating. To be a measurement, it would have to answer three questions: who answered, about which case, and what the municipality did in that case.
Without those three answers, satisfaction is no use for deciding. It is useful for the slide.
Mexico has a serious measurement of satisfaction with public services. INEGI's National Survey of Government Quality and Impact (ENCIG) 2025, published on May 21, 2026, surveys the population aged 18 and over living in localities of at least 100,000 inhabitants. Its results leave little room for doubt:
- 47.2% said they were satisfied with basic and on-demand public services. In 2023 it was 50.0%.
- Among basic services, 71.5% are satisfied with garbage collection and 51.4% with drinking water.
- Satisfaction drops to 43.4% for drainage and sewerage, 42.0% for parks and public gardens and 41.1% for street lighting.
- Only 23.7% are satisfied with streets and avenues, the lowest level among the basic services reported in the press release.
The same survey records another figure worth reading alongside the previous ones: 82.3% of those who carried out procedures, payments or service requests in 2025 said they were satisfied with that experience overall.
The two figures do not contradict each other. They measure different things. A person can leave satisfied from the counter where they were treated well and remain dissatisfied with the street they live on. The first is the experience of a contact; the second is the state of a service.
ENCIG is indispensable for knowing where one state stands against another. The press release ranks the states from Coahuila, at 58.3%, to Guerrero, at 36.5%. But a public services director cannot run operations with it: it is conducted every two years, its results are presented by state and by type of service, and it does not say which neighborhood, which crew or which report explains the result. It says how the service is doing. It does not say which case to handle differently tomorrow.
That takes another measurement, smaller and far more frequent: the one done case by case.
The Organisation for Economic Co-operation and Development (OECD) devotes a section of its Government at a Glance 2025 report to the measurement of administrative services. It starts from its Recommendation on Human-Centred Public Administrative Services, which, according to the report, encourages countries to measure performance, systematically gather feedback from users and use it for continuous improvement.
The data comes from a survey of officials responsible for services, collected between November 2024 and March 2025:
- 14 of 27 countries (52%) run national user experience surveys. Another 6 have no national survey but do run surveys at the ministry or agency level. The report cites Chile's Measuring User Satisfaction Survey (MESU), conducted annually, as an example.
- 17 of 30 (57%) use that data to publish service performance reports and communicate results.
- 16 of 30 (53%) incorporate it into their decision-making to improve services.
- Only 9 of 30 (30%) use it to measure the impact of their reforms.
The last part is the one that matters. Almost half of the governments that reported do not say they use what they measure to decide. Asking is the easy part. The hard part is getting the answer to someone who can change something.
New York shows what case-by-case measurement looks like. According to the city, when an agency closes a 311 service request, a survey is sent to the person who filed it, to learn how satisfied they were with the agency's handling of the request. Results are shared with the agencies and published on a dashboard.
The open dataset for that survey reveals three design decisions:
- Every response is tied to its request. Each row contains the response and information about the service request it corresponds to.
- Resolution is separated from treatment. The city has another, separate survey to evaluate the people who answer 311 calls. One measures whether the problem was resolved; the other, how the service was.
- The respondent is protected. To preserve anonymity, the time of the response is published aggregated to the month and the location, to the borough.
There is a fourth, less comfortable feature: the survey is offered to those who provided their contact information. Those who reported without providing it are not included. Every case-level survey has a limit like that, and the honest thing is to know it.
An average star rating can be high and badly measured at the same time. The flaws are almost never in the question, but in everything around it.
Nobody knows how many people were asked. A 4.6 from 40 responses out of 4,000 closed tickets does not describe the municipality's service. It describes 40 people. Without the number of surveys sent and answered next to the average, the figure cannot be interpreted.
The people who answer are the ones with an experience they want to share. Someone who ended up very happy or very upset has more reason to respond than someone in between. And those who have stopped expecting anything from the municipality rarely open the link. Silence is not satisfaction.
The response is not tied to a case. A generic survey does not distinguish between the pothole, the streetlight and the paperwork. If the rating drops, nobody knows which department to review. If it rises, nobody knows what was done well.
Only what was closed gets surveyed. If the survey opens upon resolution, expired tickets and those still open have no say. A department can have an excellent rating among the few cases it resolves and a backlog nobody evaluated.
Resolved for the system is not resolved for the resident. The crew marks the ticket as handled and the resident still sees the leak. We have already explained why closing the loop requires evidence, not just status. The survey is the only part of the loop where the one confirming the closure is not the municipality.
Nobody reads the comments. Free text is where residents say what the scale does not capture: "they came, but they left the rubble." If it is stored in a table nobody opens, that information does not exist.
No statistics department is needed. What is needed is for the survey to be part of the case and not a separate form.
1. One survey per ticket, not one general survey. Every response should be readable next to its case: report type, responsible department, dates and evidence. That way, a low rating leads straight to what happened.
2. It opens when the case is resolved. The natural moment to ask is at closure, through the same channel where the resident already receives their notifications. The more steps between the notification and the question, the fewer people answer.
3. First ask whether it was resolved. Before the stars comes a yes or no question: was the problem resolved? A "no" is not a bad rating; it is a case that needs to be reviewed again.
4. Report the average with its denominator. Surveys received, average rating and date of the last response go together. An average without a number of responses should not reach a meeting.
5. Read it next to what the survey does not see. On-time compliance, reopenings and backlog tell the part of the story that does not answer surveys. A department with a good rating and many expired tickets is not doing well; it is measuring only what went well.
These rules have a consequence worth anticipating: the rating usually drops when you start measuring well. Asking everyone who had a ticket resolved, and not only those who ended a conversation happy, yields a less flattering and more useful number. We have already written that an acknowledgment is not an answer; a star is not a resolution either.
Agora, Intello's citizen services platform, treats the survey as part of the ticket and not as a standalone form.
When a ticket moves to Resolved, its survey opens, and the WhatsApp notification the resident receives carries the link to answer it. In the citizen portal, where every resident checks their tickets without creating an account, the resolved ticket shows the button to respond. The question arrives through the same channel and at the same moment they are told their case was closed.
In WhatsApp follow-up, the assistant reports every change to the ticket, asks whether it was really resolved and asks the resident to rate the service. Surveys support 1 to 5 stars, Yes/No questions and free comments, so the confirmation of closure and the rating can be kept separate.
The team sees surveys received, the average rating and the last response, with averages and distribution by question. It can filter by date and, in the central office, by department. Each survey opens in detail with its ticket and its answers, so a low rating leads to the specific case, with its change log and its timeline.
Satisfaction is not read alone. Agora's analytics brings together on the same platform tickets about to expire and expired, the percentage handled on time, resolution times, reopenings and backlog by department. That lets the mayor's office compare report volume, on-time compliance and satisfaction for each department on the same dashboard, and know which departments respond and which need support.
- Can you say how many surveys were sent and how many were answered behind your average rating?
- Is every response tied to the ticket, the report type and the department that handled it?
- Does your survey ask whether the problem was resolved, in addition to asking for a rating?
- What happens when someone answers that their case was not resolved? Does anyone review it?
- Do you read satisfaction alongside each department's on-time compliance, reopenings and backlog?
- Does anyone read the free comments every week?
If several answers are no, your municipality has a rating, but it does not yet have a measurement.
Citizen satisfaction is the only service indicator the government does not produce itself. That is why it is worth so much and why it is so easy to spoil: it is enough to ask only those you want to hear from, or to separate the response from the case that explains it.
ENCIG tells a municipality how its population sees it every two years. The per-ticket survey tells it every week which case went wrong, in which department and why. Both are needed, but only the second lets it correct course before the next measurement.
If your municipality wants to measure satisfaction ticket by ticket and read it alongside each department's compliance, see how Agora works, review how Tribuna brings the same traceability to public safety and municipal justice, or request a demo.
- INEGI, Press release 28/26, Encuesta Nacional de Calidad e Impacto Gubernamental (ENCIG) 2025 (May 21, 2026)
- OECD, Government at a Glance 2025, section 4.8, "Measurement, engagement and improvement of public administrative services" (June 19, 2025)
- City of New York, NYC311, Resolution Satisfaction
- NYC Open Data, 311 Resolution Satisfaction Survey