Customer satisfaction interview questions: get behind the score

You have a score. It can be a satisfaction rating, a star rating or a would-you-recommend question, and it comes with a few comments: "great service", "fine", "a bit slow". Someone asks what to change, and the number has no answer. It tells you where people landed. It does not tell you how they got there.

The way to find out is to phone a small, deliberately chosen group of customers and ask each one to walk you through a recent visit. This guide gives you three things: a who-to-call grid built from score bands, one interview guide with three branches (positive, mixed and negative experiences) that share an opening and a close, and a way to read your notes without pretending twelve calls are a survey.

Two rules carry most of the weight. Start from what happened, not from the number. And never ask anyone to defend their score.

Everything here applies whether the people behind the score are customers, users, patients, members, or your own staff rating an internal service.

Interviews explain; surveys count

Use each method for the job it does well. Qualitative methods suit questions about why something happened and how to fix it. Quantitative methods suit questions about how many and how much.¹ Your survey has already done the counting. Interviews answer what it cannot: what the person went through, what they expected, and what it led them to do.

It follows that a handful of interviews cannot tell you what share of customers feel something. "Four of the twelve people we spoke to mentioned rebooking" is a true sentence. "Most of our customers struggle to rebook" is not. The statistical standard for generalising from a sample to a population does not apply to qualitative work, so it is the wrong yardstick for interviews.² Judge your interviews by whether they explain something you can act on. Then use a short survey question to count it. The worked example below shows that handover.

Who to call: pick by score band, and include the middle

Do not phone only the angriest customers and the happiest ones. The middle is easy to skip, and it is the first place to look for small, fixable friction: people who were neither upset enough to complain nor pleased enough to say why.

Cut your own scale into three bands and write the cut-offs down before you look at names. On a five-star scale that could be five stars, three or four stars, and one or two stars. Then plan a first round of twelve calls:

Three rules for filling the grid.

• Recent. Choose people who used you in the last few weeks, so the story is still fresh.
• Spread. Cover your sites, services and staff, not one corner. Sampling for the widest range of cases is a standard way to document diverse variations and the patterns that hold across them.³
• Not only the people who answered. People who answer a survey can differ from people who do not, a gap known as nonresponse bias.⁴ That is why the grid keeps a seat for someone who stayed silent.

Invite people in writing first. Follow your organisation's consent and privacy rules for how you contact and record people.

Hi [first name], thanks for using [organisation] recently. I'm [name]. I'd like to ask you about what your last visit was like, in your own words. It takes about 20 minutes on the phone, at a time that suits you. I'm not selling anything, and there is no right answer. Would you be willing to talk? If so, tell me two or three times that work.

The interview guide

One guide, three branches. The opening, story, branch and close below can be copied straight into a call sheet. The opening and the close are identical for everyone. In the middle you choose a branch according to the story the person told you, not the number they gave. A five-star customer can tell you a mixed story, and a three-star customer can tell you a glowing one.

Plan for 20 minutes: three for the opening, eight for the story, seven for the branch, two for the close.

The shared opening (3 minutes)

Say this, or something close to it:

Thanks for making time. I'm [name] from [organisation]. I'm not here to sell anything. I want to understand what it was like, in your words. If something was poor, telling me is useful. Is it all right if I take notes?

Then settle the permission to record or take notes in line with your organisation's rules. Do not mention their score yet.

The shared story (8 minutes)

Interview guidance advises focusing on stories and real examples and avoiding generalities.⁵ So the first real question is about one occasion, not about their opinion of you.

1. "Tell me about the last time you used us. Start from when you first decided you needed us."
2. Then, as they talk: "What happened next?" "Who was involved?" "What were you thinking at that point?"
3. After the story, not before: "Before it started, what were you expecting?"
4. Last in this block: "You rated us [score]. When you picked that, which moment were you thinking of?" If they sent no score: "You didn't send a rating back. If you had, which moment would have decided it?"

Stay quiet as much as you can; the story is theirs to tell. Notice that question 4 asks which moment, not why that number. "Why not a five?" asks for a defence. "Which moment?" points them back at something that happened.

Choose a branch by the story

Listen to the story and sort it:

• Positive: the experience went well from start to finish.
• Mixed: there were good moments and weak ones.
• Negative: something went wrong and stayed wrong.

If you cannot tell, use the mixed branch. It is the safest.

Branch A: positive experiences (7 minutes)

You are looking for what to protect, and whether it depends on one person.

1. "Take me to the moment that made it good. What exactly happened?"
2. "Who did what?"
3. "What would it have been like if that had not happened?"
4. "Would it have been as good with someone else on duty?"
5. "What did it lead you to do afterwards? Come back, tell someone, nothing?"
6. "Was there any moment that was less good, even a small one?"

Branch B: mixed experiences (7 minutes)

Do not skip this branch.

1. "Take me through the good part and the not-so-good part. Which came first?"
2. "Which of the two has stayed with you more?"
3. "At the weaker moment, what did you expect to happen? What happened instead?"
4. "What did you do then?"
5. "How did you feel about us when you left?"
6. "If we could change one thing first, what would it be?"

Branch C: negative experiences (7 minutes)

Do not defend, explain or promise anything. Your job is to understand.

1. "Take me back to the point it went wrong. What happened just before?"
2. "What were you expecting at that point?"
3. "What did you do then? Did you tell anyone?"
4. "What did we do in response, if anything?"
5. "What would have put it right for you?"
6. "What has it led you to do since?"

If the person needs a problem solved, say you will pass it on to your complaints route, and then do it straight after the call. Do not try to fix it live.

The shared close (2 minutes)

1. "Is there anything I should have asked and didn't?"
2. Tell them what happens next: you are talking to a small number of people, the notes are used to improve the service, and any individual problem has gone to the right place.
3. Thank them.

For question craft beyond this guide, such as what to ask when the first answer is vague, see how to ask customer interview questions that move beyond surface answers.

Pull each story apart

After every call, write the notes the same day, using one sheet per call. The sheet separates three things that satisfaction scores blur together: what happened, what the person expected, and what it led them to do.

CALL NOTES
Call ID:                    Date:
Site or service:
Score given (or none):
Branch used: positive / mixed / negative

The moment (their words, in order):

What they expected:

What happened:

What it led them to do:

The change they would make:

A surprise (what I did not expect):

Put the person's own words, in quotation marks, in the first box. The other boxes can be your summary.

Write the expectation in its own box, so you can see whether a low score came from the service or from the expectation, and a high one from the service or from low expectations.

"What it led them to do" matters most when you decide what to change. A customer who was mildly annoyed and rebooked tomorrow is a different case from one who was mildly annoyed and quietly stopped coming.

When the sheets are done, lay them side by side and list the moments that appear more than once. Write "four of the twelve people we spoke to", never a percentage of customers. To turn the repeated moments into claims you can defend, use the checks in what makes a customer insight evidence-backed.

An invented example: a three-site physiotherapy clinic

Here is an invented example to show the method. The clinic, the people and the numbers are made up.

A physiotherapy clinic with three sites sends a five-star rating request when a patient's course of treatment ends or their file is closed. The average has sat at the same level for months. The manager wants to know what to change. She plans the twelve-call round from the grid, spread across the three sites and the clinic's clinicians.

Notice what the stories did to the bands. One of the three five-star patients told a mixed story: the treatment was excellent, but she waited a long time to rebook. One of the five middle patients had a glowing story and had simply never given a five in his life. The caller used the branch the story called for each time, not the one the band suggested.

Call 6, from the middle band, shows the note sheet filled in:

CALL NOTES
Call ID: 6                  Date: [date]
Site or service: Site B, knee rehabilitation
Score given (or none): 3 stars
Branch used: mixed

The moment (their words, in order):
"I liked her, but I got to the end and had no idea what
to do at home between sessions."

What they expected:
A clear plan to follow.

What happened:
Was told "keep doing the exercises". Nothing in writing.

What it led them to do:
Skipped some days. Left it longer than told before returning.

The change they would make:
A printed sheet with the exercises.

A surprise (what I did not expect):
Gave 3 stars for the sheet, not the treatment.

Reading across all twelve sheets, two moments repeat:

• Booking and rebooking. Four of the twelve people described friction here (calls 3, 4, 5 and 12). Call 12, the person who gave no score, liked her clinician but stopped coming after the third session because she could not get an evening slot.
• The home exercise sheet. Four of the twelve mentioned it. Calls 1 and 2 praised a printed sheet with photos. Calls 6 and 8 never received a clear one. Calls 1 and 2 saw the same clinician; calls 6 and 8 saw two others. The same moment sits in good stories and in mixed ones, which suggests the difference may lie with who hands the sheet out.

The three negative calls told three different stories. Call 9 was cancelled by the clinic on the day. Call 10 saw a different clinician each time. Call 11 sent a message about worsening pain and nobody replied. The caller passed all three to the complaints route the same day, as the guide says. One call is not a pattern, but each one describes a broken promise, so the manager also checks the records for all three.

She makes three decisions. Two are changes, framed as tests with the decision rule agreed in advance, and Site C carries on as now to act as the comparison. She adds the two survey questions below to the rating request before the trials begin, so each site has a before-figure. Each change applies to every patient who starts in that month and stays with them until they finish. She reads the gap between sessions after the month, and the survey answers when those patients finish their course.

1. Site A tries booking the next appointment before the patient leaves the room, for a month. She compares each site's gap between sessions with its own previous month, then Site A's change with Site C's, and reads the follow-up appointment survey answer the same way. If Site A improved no more than Site C, she drops the change.
2. Site B tries the same printed exercise sheet for every clinician, for a month. She compares Site B's change in the survey answer about knowing what to do at home, from its own before-figure, with Site C's. If Site B improved no more than Site C, she drops the change.
3. She checks the records for calls 9, 10 and 11. This is a check, not a change.

The two questions are: "During your treatment, could you always get a follow-up appointment at a time that suited you?" and "Between sessions, did you know what to do at home?" Both suit a request sent when a course ends. The interviews found the moments. The survey questions count them.

What this will not tell you

This guide will not tell you how many customers feel a given way; twelve calls cannot, and that is a survey question's job. It will not tell you what people who refuse a call think. Treat the people who agree to talk as one part of the picture, not a map of everyone.

Recall has its own limits. Methods that ask people to recall a notable incident often bring back extreme events, while small issues and typical use are rarely mentioned.⁶ That is why you also read your survey comments, your complaint logs and your booking data. And a call can show a plausible reason for a score, but it cannot prove that fixing that reason will raise the number. Treat each change as a test.

Next step

Pick your three bands and write the cut-offs down. Choose the first three people, one from each band, and send the invitation this week. Run those three calls with the guide, fix whatever felt awkward, and then complete the other nine. If you would like to see how this kind of follow-up study is set up as a whole, there is an overview of customer satisfaction research.