Most companies already collect customer feedback. The harder part is getting answers you can trust and turning them into a decision someone acts on. This guide is the map for the whole topic: which survey to run, how to write the questions, how to get people to respond, how to get honest answers, and what to do with the results. Each section points to a deeper article when you want the detail.
Key takeaways
- Customer surveys measure satisfaction, loyalty and needs, and they produce decisions when each is designed around a specific business question and kept short.
- Choose the survey type before writing questions: relationship surveys run on a schedule, transactional surveys follow an interaction, and CSAT, NPS and CES are the metrics they carry.
- Keep the survey under ten minutes, or five on mobile, and ask about specific recent behavior to get answers people can give honestly.
- Add one open-ended question after the key rating, such as the main reason for the score, and read it alongside the numbers.
- Cut results by segment before concluding anything: a 70% satisfaction average can hide a 36-point gap between existing and new customers.
What are the main types of customer surveys?
Surveys differ by when you send them and what they measure. Choosing the type first keeps the questionnaire short, because each type answers one kind of question.
| Survey type | When you send it | What it tells you | Typical question |
|---|---|---|---|
| Relationship survey | On a schedule, such as quarterly or yearly | How customers feel about the company overall | ”How likely are you to recommend us?” |
| Transactional survey | Right after a specific interaction, such as a purchase or support call | Whether that moment worked | ”How satisfied were you with today’s service?” |
| CSAT | After an interaction or on a schedule | Satisfaction with a product, service or event | ”How satisfied are you with…?” (1 to 5) |
| NPS | Usually in relationship surveys, sometimes after interactions | Loyalty and likelihood to recommend | ”How likely are you to recommend us?” (0 to 10) |
| CES | After a task, such as resolving an issue | How easy the experience was | ”The company made it easy to handle my request.” (agree to disagree) |
Relationship and transactional describe the timing. CSAT, NPS and CES describe the metric. You can run a transactional CSAT after a support call and a relationship NPS twice a year, and they will answer different questions.
What is a customer satisfaction (CSAT) survey?
A CSAT survey asks customers to rate how satisfied they were, usually on a five-point scale, and reports the share who chose the top two points. It is the most common customer survey because it is simple and can be tied to one moment. Our article on what a customer satisfaction survey is and why it matters covers the question wording and when to send it.
What is Net Promoter Score?
Net Promoter Score asks one question on a 0 to 10 scale: how likely are you to recommend us to a friend or colleague? Customers scoring 9 or 10 are promoters, 7 or 8 are passives and 0 to 6 are detractors. NPS is the percentage of promoters minus the percentage of detractors. The full method, with a worked calculation, is in what is Net Promoter Score.
What is Customer Effort Score?
CES asks how easy it was to get something done, using a statement customers agree or disagree with. It suits service and support, where a smooth fix matters more than a delightful one. Use it after a task, not as an overall loyalty measure.
How do you design a survey that produces decisions?
Start from the decision, not the questionnaire. Write down what the business will do differently depending on the answer. If nothing would change, cut the question.
- Name the decision. “Should we change the onboarding emails?” is a decision. “Learn about customers” is not.
- Pick the survey type. Match the decision to relationship or transactional timing and a single headline metric.
- Ask about specific, recent behavior. “Think about your last order” beats “In general, how do you feel about ordering from us?”
- Keep scales consistent. Use the same number of points and the same direction throughout the survey and across waves, so movement over time means something.
- Keep it short. As short as the objectives allow. Under ten minutes is a sensible ceiling for most customer surveys, and under five on mobile.
- Test before launch. Have a few people take it on a phone, then read their answers. Misread questions show up quickly.
The same logic applies when you do not yet know what to ask. If your scores look fine but you cannot say what customers want, you may be experiencing customer disconnection, which is a good reason to survey on a regular schedule rather than only after a problem.
Should you use open-ended or closed questions?
Use both, for different jobs. Closed questions, where people pick from a list or a scale, give you numbers you can compare across segments and over time. Open-ended questions let customers say what you did not think to ask, in their own words.
The practical rule is one well-placed open-ended question, usually right after the key rating: “What is the main reason for your score?” Several open-ended questions tire respondents and leave you with a pile of text to code. The benefits and challenges of open-ended survey questions goes through the trade-offs, including how to code the answers.
How should Likert and rating scales be used?
Use the same number of points throughout, label every point or at least the two ends and the middle, and keep the direction consistent. Five- and seven-point scales are common for agreement and satisfaction; the 0 to 10 scale belongs to NPS-style questions. When you report, show the distribution and the top-box percentage, not only the mean, and always show the base, meaning how many people answered.
How do you get people to respond?
The share of people who complete your survey is its response rate, and it decides whether your results describe your customers or only the most motivated ones. A low rate is a warning rather than a verdict, but it tells you to check who is missing before you trust the average. What a survey response rate is and why it matters explains how to calculate it and what counts as a respectable figure.
Four things move response more than anything else:
- Length. Every extra screen costs you completes.
- Timing. Ask close to the experience, while the customer still remembers it.
- Device. Many people open the invitation on a phone, so design for a small screen first.
- A clear reason. Say what you will do with the answers. People are more willing to help when they believe someone will read what they write.
That last point connects to why people answer at all. Many customers like being asked for their opinion, and the psychology of surveying explains why, and how to use it without pressuring anyone.
Why do people answer inaccurately, and what reduces it?
People want to look reasonable, they tire, they misread, and they describe who they intend to be rather than what they did. None of this is malicious. It is how people behave when answering questions, and good design reduces it. Why people lie on customer surveys and how to minimize it covers the causes in depth.
The levers that work best:
- Ask about the last purchase rather than typical behavior.
- Place sensitive questions late, after the respondent has settled in.
- Offer balanced answer options, with as many negative choices as positive ones.
- Avoid wording that hints at a preferred answer, such as “How much did you enjoy…?”
- Keep the survey short. Length is the lever you control most directly, and fatigue shows up as skipped questions and straight-lining.
How do you analyze survey results and act on them?
Cut by segment before you draw conclusions. A single overall score hides the groups that are doing well and the ones that are not. Then pair the scores with the verbatim themes, and track the same questions over time so movement is visible.
A worked example
Suppose a transactional CSAT survey gets 200 responses on a five-point scale (illustrative numbers). The share choosing 4 or 5, the top-two box, is 70%.
| Rating | Responses | Share |
|---|---|---|
| 5 (very satisfied) | 80 | 40% |
| 4 | 60 | 30% |
| 3 | 30 | 15% |
| 2 | 20 | 10% |
| 1 (very dissatisfied) | 10 | 5% |
| Total | 200 | 100% |
Top-two box: (80 + 60) / 200 = 140 / 200 = 70%. The mean is (5 x 80 + 4 x 60 + 3 x 30 + 2 x 20 + 1 x 10) / 200 = 780 / 200 = 3.9.
Seventy percent looks healthy. Now cut it by customer group, with 100 responses each (also illustrative):
| Segment | Responses | Chose 4 or 5 | Top-two box |
|---|---|---|---|
| Existing customers | 100 | 88 | 88% |
| New customers | 100 | 52 | 52% |
| All respondents | 200 | 140 | 70% |
The 70% average hides a 36-point gap. The decision is not “improve satisfaction” but “find out what goes wrong in the first weeks.” The open-ended answers from new customers tell you what; the segment cut tells you where to look.
How do you close the loop?
Route each finding to the team that can act on it, with a recommended action. Then measure whether the action moved the score at the next wave. mTab Pulse monitors trackers and studies and surfaces meaningful movement, so you see that change without hunting for it. Companies that do this consistently tend to treat surveying as a program rather than a one-off project, and three examples of companies surveying their customers well show what that looks like in practice.
What are the most common customer survey mistakes?
- Asking questions nobody will act on. Every question should map to a decision.
- Reporting only the average. Means hide distribution and segment gaps, as the example above shows.
- Changing scales between waves. You lose the ability to compare over time.
- Treating a low response rate as someone else’s problem. Check who is missing before you trust the result.
- Counting open-ended answers but not reading them. A handful of specific complaints can point to a fix more precisely than a theme frequency does.
How do you do this in practice?
In Halo Reports, customer surveys and satisfaction trackers are analyzed by any segment with crosstabs and banner tables, weighting and significance testing, so a gap like the one above is visible at a glance and you can tell whether it is real. Results can be shared as interactive report pages and charts, and Net Promoter Score views sit alongside the other measures. See Halo Reports for the details. If you also want to see what drives satisfaction and loyalty by segment and journey, that page shows how.
Where should you start?
Pick one decision your team needs to make this quarter, choose the single survey type that informs it, and write no more than eight questions, ending with one open-ended “why.” Run it, cut the results by segment, and put one recommended action in front of the person who owns the fix. Then read the supporting articles above for whichever part, such as response rate or honest answers, turned out to be your weak point.