What Is NPS? Recommendation Intent, the Formula, and Comparison Limits
Published:
Last Updated:
Category: Marketing Glossary
Authors: Shusaku Yosa
Net Promoter Score, or NPS, summarizes answers about willingness to recommend a company, product, or service. It does not count actual referrals or registrations. Calculate it consistently, then distinguish the stated intention from the customer experience behind it and from any behavior you can independently observe.
Understand what the question measures
NPS uses a recommendation-likelihood question on a scale from zero to ten. Bain’s official explanation classifies scores of 9–10 as promoters, 7–8 as passives, and 0–6 as detractors. The score subtracts the detractor percentage from the promoter percentage.
These are scoring labels, not conclusive descriptions of every individual. A respondent who selects nine has not necessarily made a referral. A respondent who selects six is not certain to leave. Reasons, subsequent behavior, and the scope of the customer relationship require additional information.
Include passives in the denominator
Response | Classification | Calculation treatment |
|---|---|---|
9–10 | Promoter | Added in the numerator; included in the total |
7–8 | Passive | Not added or subtracted; included in the total |
0–6 | Detractor | Subtracted in the numerator; included in the total |
NPS = (promoters − detractors) / total valid responses × 100. Saying that passives are excluded without explaining the denominator invites an error. They do not contribute to the numerator difference, but they remain part of the valid-response total.
Define valid responses before calculating. Blank, out-of-range, or duplicate records need consistent handling. Record exclusions and their reasons rather than selecting whichever responses make the score look better.
Check the arithmetic with 100 hypothetical responses
This is a fictional calculation example, not a company result. There are 40 promoters, 35 passives, and 25 detractors among 100 valid respondents. The promoter share is 40%, the detractor share is 25%, and NPS is 15.
Calculation | Result | Interpretation |
|---|---|---|
(40 − 25) / 100 × 100 | 15 | Correct denominator: all 100 valid responses |
(40 − 25) / 65 × 100 | About 23.08 | Incorrectly excludes the 35 passives |
Every response is a promoter | 100 | Upper endpoint |
Every response is a detractor | −100 | Lower endpoint |
Report NPS as a score, normally without a percent sign. If rounding is needed, apply a consistent display rule and avoid repeated rounding during aggregation. A score of 15 does not mean that 15% made referrals or that satisfaction is 15%.
Choose other measures for other questions
One number cannot describe every part of customer experience. You may want to know about a recent support interaction, the effort of completing a task, continued use, or actual advocacy. These are different questions, with different timing and data requirements.
Question | Possible measure or record | Definition to establish |
|---|---|---|
Would someone recommend us? | NPS | Question, scale, and valid responses |
Was an interaction satisfactory? | A CSAT-style question | Experience and timing |
Was a task difficult? | A CES-style effort or ease question | Wording and scale |
Did someone stay or refer? | Retention or referral records | Population, period, and observed action |
Do not rank these scores directly when their scales differ. If NPS rises while usage does not, check the populations and timing before calling the results contradictory. Respondents may not represent all users, and intentions and behavior need not move together.
Record comparison conditions and response bias
A positive score means promoters outnumber detractors as a share of valid responses. It does not establish a universal passing mark. Industry, country, language, customer mix, wording, contact point, collection method, and timing can change the interpretation.
A large response count does not remove selection bias. Instead of applying a universal minimum such as 400, report the valid count, invitations, response rate, composition, and changes from the previous wave. Sampling precision and nonresponse are different issues. Small movements should not be treated as proven improvement without considering uncertainty and comparability.
Use reasons to improve the experience
Ask why respondents chose their answers, then identify specific difficulties and strengths. Provide appropriate help where there is a problem. A high score does not mean that nothing can improve. Avoid arrangements that reward favorable scores or encourage staff to pressure customers into a preferred answer.
- Keep a record of whom you asked, when, and about which experience.
- Separate confirmed issues from interpretations of comments.
- Assign an owner and review date to the improvement.
- Check later feedback alongside independently observed behavior.
See what an advocate is for actual advocacy roles, supporting voluntary advocacy for appropriate interventions, and customer-experience value for designing improvements. Use the score as one input to a clearer understanding of customers.
