What questions should an NPS survey ask?
Start with a 0–10 recommendation question and a neutral open-ended question about the reason for the rating. Add other questions only when they serve a clear decision. A short survey is a useful goal, but there is no universal rule that every NPS survey must contain exactly two or three questions.
The rating supports the Net Promoter Score calculation. The explanation adds the respondent’s perspective. Neither automatically identifies the cause of a score change, and a blank comment should not be filled in with an assumed reason.
Use the examples below as a starting point for your audience. Decide what you will do with each answer before sending the survey, and test the wording with people who resemble the intended respondents.
A short NPS survey template
Scroll horizontally to see all columns →
| Question | Response format | Purpose |
|---|---|---|
| How likely are you to recommend [company or service] to a friend or colleague? | 0–10, with clear endpoint labels such as “Not at all likely” and “Extremely likely.” | Recommendation rating. |
| What is the main reason for your rating? | Open text. | The respondent’s explanation. |
| What, if anything, could we improve? | Optional open text, if needed. | A practical suggestion beyond the initial explanation. |
| May our team contact you about this feedback? | Appropriate contact choice, when named follow-up is part of the design. | Clarifies the follow-up arrangement. |
The first two questions form the core example. Do not automatically include the others. If the reason question already supplies the necessary improvement information, a second open question may be repetitive. Contact arrangements also depend on how the survey is distributed and what respondents have been told.
For scoring, 9–10 ratings are promoters, 7–8 passives and 0–6 detractors. Subtract the detractor percentage from the promoter percentage using all valid ratings as the denominator. See Bain’s scoring guidance.
Choose the right object of the recommendation
Make it clear what the respondent is rating. A company, a product and a recent support interaction are different objects. Asking about “us” when several services are involved can leave the team unsure which experience the answer describes.
For a B2B service, a relevant colleague may be the natural reference. For a consumer product, a friend may fit. Choose wording that makes sense, then record it and use it consistently for the intended trend. If recommendation is not meaningful for the audience, consider whether another question better serves the decision.
Scroll horizontally to see all columns →
| Situation | Design question | Useful follow-up |
|---|---|---|
| Overall customer relationship | Is the respondent familiar enough with the service to assess it? | What is the main reason for your rating? |
| New customer onboarding | Are you measuring the wider relationship or a specific setup experience? | What part of getting started most influenced your answer? |
| B2B account feedback | Which role does the contact have, and what part of the service do they use? | What most influences your view of the service in your role? |
| Member or association feedback | Is recommending membership relevant to this audience? | What most influenced your answer about membership? |
These adaptations are not interchangeable benchmarks. Keep the survey purpose visible in the analysis. A rating about one recent event should not quietly become the organization’s overall relationship score.
Write follow-ups that do not supply the answer
“Why were you unhappy?” assumes a negative experience. “What did you love?” assumes a positive one. “What is the main reason for your rating?” gives respondents room to describe an experience that may not match the team’s expectation.
A score of 6 does not necessarily mean the person disliked everything; a score of 10 does not mean there are no problems. Avoid a branch that prevents a promoter from reporting an issue or tells a detractor what the issue must be.
If you use score-based follow-ups, preserve their wording in the data. Different questions can elicit different kinds of comments, so a theme comparison across branches may reflect both experience and question design. A common reason question can make the initial comparison easier.
When should you use open or closed questions?
Open questions can reveal issues the team did not anticipate and preserve the respondent’s language. Closed questions can efficiently measure defined topics across respondents. Neither format is universally better; they answer different measurement needs.
For example, an open question can discover unexpected onboarding problems. A later, carefully tested question about a known setup step can measure how often a specific difficulty occurs. Do not treat the percentage who spontaneously mention a topic as equivalent to the percentage who select it when asked directly.
Pew Research Center describes how response formats, wording and question order can influence answers, and why pretesting matters. See its question-design guidance. Use that principle to choose each item rather than assuming automated text analysis makes closed questions unnecessary.
For a closed follow-up, supply relevant, understandable options and an appropriate way to indicate that none applies. Avoid overlapping choices unless the format deliberately allows several answers. Do not add a long list of topics simply because the survey tool makes it easy.
How many questions are enough?
Keep the questionnaire proportional to the decision and the respondent’s effort. Review every proposed item: what will change because of this answer, who will use it and can the information be obtained appropriately without asking again?
A single open question can require more thought than several simple selections. Assess the actual experience, including mobile typing, language and accessibility, rather than counting items alone. Do not promise a higher response rate merely because a questionnaire is shorter.
Use a pilot to examine completion, skipped items and how people interpret the questions. If responses to two open questions repeat each other, consider combining them. If the answers lack information necessary for action, improve the wording or add a focused question.
Preserve comparability while allowing useful changes
Keep the intended trend question, scale and context stable where possible. Record the collection channel, audience, timing and question order as well as the text. A score can move when collection conditions change, even if the underlying experience does not.
Sometimes a question should change because it is unclear or no longer fits the service. Document the old and new versions and assess the implications. Depending on the change, a parallel test, a marked break or a qualified comparison may be appropriate. Do not assume every small edit invalidates all history, or that every revision is harmless.
Across products or locations, agree on a small common core where comparison is needed. Allow relevant local follow-up questions and use a data dictionary to record their meaning. Keep stable customer context separate from recurring answers, and update changing attributes when necessary.
Explain identity and follow-up clearly
Decide whether responses are anonymous, confidential within a defined team or connected to a named customer relationship. Explain the actual arrangement. A personalized link or imported account information may affect whether a promise of anonymity is accurate.
For named follow-up, collect or use only the information needed for that purpose. For anonymous research, retain the arrangement when reviewing comments and sharing results. Avoid asking respondents to include private details in an open field when those details are unnecessary.
In a B2B program, a person and an account are different units. Several contacts may describe different experiences at one account. Preserve the role and account relationship where appropriate, and decide how any account-level reporting will be constructed.
Example: plan the analysis before sending
Imagine a fictional product team asking 500 customers the core rating and reason questions. One hundred give valid ratings: 50 promoters, 30 passives and 20 detractors. The NPS is 30 and the response rate is 20% under that invitation denominator.
Sixty provide usable comments. If 12 of those comments mention setup, that is 20% of usable comments. It is not evidence that 20% of all 500 customers experienced a setup problem. The question asked for a main reason, so other experiences may go unmentioned.
The team should show the rating counts, comment coverage, reviewed themes and limitations together. If it needs to estimate the prevalence of a specific setup problem, that may require a more focused question and suitable collection design. The initial comments help define that next question.
Before you launch the survey
- State the decision and the audience.
- Confirm the recommendation question fits the experience being measured.
- Test wording, scale labels and mobile completion.
- Check the identity and contact explanation.
- Record the questionnaire version and collection conditions.
- Calculate a small test dataset independently.
- Assign ownership for reviewing comments and responding to issues.
Sopact’s collection, analysis and governance approach is relevant when the team needs to maintain this process across repeated cycles and sources. Verify that the proposed setup preserves definitions, appropriate context and reviewed findings. Question design still requires judgment; an automated summary does not repair an ambiguous question.
Continue to NPS survey software for a practical tool evaluation, or the detractor response guide for reviewing and acting on low ratings. The customer feedback management guide connects these activities into a broader program.
Watch: reviewing feedback after collection
This video discusses reading incoming feedback and keeping useful context available for review. Use the approach with the collection and privacy choices appropriate to your audience.
Frequently asked questions
What is a useful NPS follow-up question?
“What is the main reason for your rating?” is a neutral starting point. Test whether it provides the explanation your team needs without assuming a positive or negative experience.
Must an NPS survey have exactly two questions?
No. A rating and reason question are a useful core. Add questions when they serve a clear purpose and keep the respondent’s effort proportionate.
Are open-ended questions always better?
No. Open questions help discover and understand issues; closed questions can measure defined topics. Choose the format around the information needed and do not treat their results as interchangeable.
Can we change the NPS wording?
Yes, when there is a good reason, but document the change and assess its effect on comparison. Stable wording and collection context make trends easier to interpret.
Should the reason question be required?
Consider the purpose and test the experience. Forcing an explanation can create burden or low-information answers. Keep valid ratings and comment coverage distinct in the analysis.
Should every low rating receive a different follow-up question?
Not necessarily. A common neutral reason question allows people across the scale to describe their experience. If branching is useful, preserve the branch wording and account for it when comparing comments.
Choose your next step
King Center
Related practice: using participant feedback to inform program review. This story illustrates feedback use rather than NPS benchmarking.
Read the customer story →