How to Run an NPS Survey: Invitation Examples, Reporting, and Follow-Up
An NPS survey asks customers how likely they are to recommend a business or service on a scale from 0 to 10, then uses their reasons to guide improvement. Running one requires more than sending a question: define the experience, explain the request, count valid responses consistently, and assign follow-up work. A score does not establish what every customer thinks or prove future revenue.
Separate survey design from day-to-day execution
Decide whether you are asking about the overall relationship or a particular purchase, support interaction, or other experience. The audience and timing affect what the answer means. For question and audience design, see how to design an NPS survey.
Before sending, test duplicate suppression, opt-out handling, links, mobile display, and the completion screen. Estimate completion time from a test rather than a guess. Participation should be voluntary. Do not imply that declining or giving a low score will affect service.
Write an invitation that explains the request
Element | Include | Avoid |
|---|---|---|
Purpose and experience | The service and experience you want to improve | A vague request about everything |
Effort and deadline | Tested time estimate and closing date | An unsupported “takes one minute” promise |
Use and choice | Voluntary participation and how answers are used | Calling identifiable responses anonymous |
Contact | Sender and a working help contact | Sales pressure or requests for high scores |
Here is a fictional invitation to adapt: “Thank you for using Service A. We would appreciate your feedback about your recent experience so we can improve the service. Participation is voluntary. Please respond at [link] by [date]. The survey takes approximately [time measured in testing]. Find information about how we use responses at [notice], or contact [team] with questions.”
Exclude completed responses and opt-outs from reminders. If you offer individual follow-up, distinguish that choice from anonymous participation. Never ask respondents to help the team by selecting a favorable score.
Read NPS case studies through their conditions
A high published score is not automatically a useful target. Look for practices you can adapt, such as routing feedback to an owner. Do not invent a company’s results or assume an improvement came from the featured initiative.
Condition | Ask about the case | Why it matters |
|---|---|---|
Audience and timing | Who was invited, and after what experience? | Former or inactive customers may be missing |
Method and wording | Which channel, question, and scale? | Changes can alter responses |
Counts | Invitations, valid answers, and nonresponses? | Respondents may differ from nonrespondents |
Action and outcome | What changed, and what else changed? | A before-and-after difference is not causal proof |
AAPOR’s survey guidance explains the need to keep wording and methods comparable over time and identify missing or partial answers. Matching industry or country alone does not make two studies comparable.
Keep passives in the denominator
Bain defines promoters as scores of 9–10, passives as 7–8, and detractors as 0–6. NPS is the percentage of promoters minus the percentage of detractors. Passives remain in the valid-response denominator; they simply add nothing to the difference in the numerator. Missing answers are not zero scores.
In a fictional survey A, 500 invitations produce 100 valid answers: 60 promoters, 30 passives, and 10 detractors. The response rate is 20%, and NPS is (60 − 10) ÷ 100 × 100 = 50. Report a score of 50, not 50%. The views of the 400 nonrespondents remain unknown.
Suppose a comparable, nonoverlapping group B has 20 answers: 4 promoters, 10 passives, and 6 detractors. Its NPS is −10. The combined score is (64 − 16) ÷ 120 × 100 = 40. Averaging the two scores without weighting gives 20 and is incorrect for the pooled respondents.
Turn reasons into assigned work
Code reasons into useful categories, such as delivery, usability, or support. Record counts and the underlying problem. If one answer can have several themes, say so; theme percentages need not sum to 100%. Remove identifying details from quotations and avoid publishing combinations that expose a respondent.
- Route serious service or safety issues promptly, regardless of the score.
- Prioritize recurring issues by frequency, impact, and effort.
- Record an owner, deadline, action, and review date.
- Repeat under comparable conditions and inspect response composition as well as the score.
For ongoing dialogue, consider community operating practices. Measure whether customers actually remain with customer retention rate. Recommendation intent and observed retention are different outcomes, and neither alone proves that your intervention caused a change.