Public services
Public service satisfaction survey
Ask about one transaction while the visitor still remembers it: what was unclear, how long it took, how many times they came back.
Why this survey gets sent
“Our satisfaction score is 81 percent. I have been asked to raise it to 85, and nobody — including me — can name one thing we would change tomorrow morning to do it.”
A score is a summary of things that happened one at a time: a requirement nobody explained, a document asked for twice, a counter that closed at noon. Ask about the transaction rather than about the institution, and the same two minutes come back as a list of repairs.
What is in it
- Which service, and through which channel
- Clarity, time, staff conduct, fair treatment — one rating grid
- Was the service completed? (single response)
- How many times you had to come back (single response)
- Were you asked for the same thing twice? (yes / no)
- The one change that would have made it easier (long answer)
About 3 min to answer
Public service satisfaction survey
About 3 min
How to write this survey
Ask about one transaction, not about the institution
The single most useful decision here is to attach the survey to one transaction. A person who renewed a licence this morning can tell you precisely what was unclear on the requirements page. The same person, asked how satisfied they are with the department in general, answers with how they feel about waiting rooms, or about the last thing they read.
That is why the first two questions are which service and which channel. They cost the respondent five seconds and they turn every later answer into something addressable: the counter is fine and the website is not, or the licence renewals are fine and the certificates are not. Without them you hold one number for an organisation that does twenty different jobs.
Measure effort, not only feeling
Satisfaction is a feeling and it drifts. Effort is a count, and it does not. “How many times did you have to come back?” and “were you asked for the same document twice?” are answered the same way by a patient person and an impatient one, which is what makes them comparable between months and between branches.
These two questions also point at a department by name. A document asked for twice is almost never rudeness; it is two systems that do not read each other, or a checklist that lives in one office and not in the other. That is a fixable thing with an owner, which is more than a satisfaction percentage has ever handed anybody.
Anonymity is what buys you the truth
Say on the first screen that you are not asking for a name or a reference number, and then do not ask for one. People who are waiting on a decision believe — reasonably — that a criticism with their name attached travels back to the person deciding, and they answer accordingly: everything good, nothing useful.
Anonymity has a price and it is worth paying: you cannot go back to one person and repair their case. Run the individual recovery through the complaints channel, which is built for it and asks for a name openly, and let this survey do the other job — telling you what is broken for everybody rather than for one. If you must offer a callback, make it a separate optional question and say plainly that using it means the response is no longer anonymous.
Write it in the visitor's words, not the department's
Every line should use the words heard at the counter. “How clear the requirements were before you started” is a sentence a visitor recognises; the internal name for the same thing — the service card, the pre-submission checklist — is a sentence only your staff recognise, and a question nobody understands is answered in the middle column by everyone.
Label every column of the grid rather than only its ends, and keep the same labels next quarter: the number only means something next to the last one. Avoid the double question — “were the staff polite and quick?” — because a visitor who met a courteous employee sitting in front of a frozen screen has no honest answer to give you.
Read it by channel, then publish what changed
Split the results by channel and by service before you look at any total. The average hides the interesting thing almost every time: the phone line that sits three points below everything else, the one service whose requirements nobody can follow. Two filters answer the question the total cannot — where do we start?
Then publish what you changed. A short notice at the counter — “you told us the requirements page was unclear; here is the new one” — does more for the next round's response rate than any reminder, because it is the only proof a visitor ever gets that answering was not a waste of their afternoon. A survey that never produces a visible change trains the people who answered it to stop.