Key takeaways
The Net Promoter Score℠ (NPS®) measures the share of customers ready to recommend a company, based on a single question rated from 0 to 10. The score is the share of promoters (ratings of 9 and 10) minus the share of detractors (ratings of 0 to 6): a number between -100 and +100. It only means something alongside its number of responses.
- Origin: created in 2003 by Fred Reichheld, a partner at Bain & Company, in the Harvard Business Review.
- Good score: there is no universal threshold. The only median published by the inventor, in 2003, was 16; Bain advises comparing yourself with your competitors.
- Reliability: with 100 responses, an NPS of 30 has a margin of error of about ± 15 points at 95% (Elev8 Lab calculation).
What is the Net Promoter Score?
The Net Promoter Score is a recommendation metric: it sums up in one number how willing customers are to recommend a company.
Definition
NPS rests on a single question: “How likely is it that you would recommend [company X] to a friend or colleague?” The answer goes from 0 (“not at all likely”) to 10 (“extremely likely”), with 5 meaning “neutral”.
This wording comes from the founding 2003 article; Bain now uses “recommend us”.
| Group | Ratings | What Bain says |
|---|---|---|
| Promoters | 9 and 10 | More than 80% of referrals in most businesses |
| Passives | 7 and 8 | Repurchase and referral rates up to 50% lower than promoters |
| Detractors | 0 to 6 | More than 80% of negative word of mouth |
These proportions are claims made by Bain, with no published method. NPS is one of the four metrics of customer loyalty, along with retention, churn and customer value. It measures an intention, not a behavior.
Where does NPS come from?
NPS was born from an article by Frederick F. Reichheld, “The One Number You Need to Grow”, published in the Harvard Business Review in December 2003.
About twenty questions were tested with more than 4,000 customers in six industries. Compared with actual purchases and referrals, the recommendation question ranked first or second in 11 of 14 case studies.
The author declares a conflict of interest: he sat on the board of directors of Satmetrix, a customer feedback software company he worked with. He also pointed out two limits as early as 2003: industries dominated by monopolies, and some B2B markets where decision makers were missing from the samples. The measure later became the Net Promoter System℠, which Bain presents as a management system.
NPS, a registered trademark
Net Promoter®, NPS® and NPS Prism® are registered trademarks of Bain & Company, NICE Systems and Fred Reichheld; Net Promoter Score℠ is a service mark of the same owners.
On its “Trademarks and Licensing” page, Bain encourages free use of the method. It requires attribution (® or ℠ symbol at the first occurrence, owners named in a note) and a license for commercial use of the trademarks.
Key point
Using the method is free. Citing the trademark requires the symbols and an attribution notice, placed here below the sources.
How do you calculate the Net Promoter Score?
The Net Promoter Score is calculated by subtracting the share of detractors from the share of promoters. Bain’s formula fits on one line: NPS = % of promoters – % of detractors.
- Ask the question: a rating from 0 to 10, with no other scale.
- Sort the responses: 9 and 10 for promoters, 7 and 8 for passives, 0 to 6 for detractors.
- Calculate the shares: each group divided by the total number of responses, passives included.
- Subtract: promoters minus detractors, with the result written without a % sign.
A fictional example by Elev8 Lab:
| Group | Ratings | Responses | Share |
|---|---|---|---|
| Promoters | 9 and 10 | 100 | 50% |
| Passives | 7 and 8 | 60 | 30% |
| Detractors | 0 to 6 | 40 | 20% |
| Total | 0 to 10 | 200 | 100% |
The NPS is 50 – 20 = 30. Passives drop out of the subtraction but stay in the total: each extra passive pulls the score toward zero. And a 6 weighs exactly as much as a 0.
What is the margin of error of an NPS?
The margin of error of an NPS depends on the number of responses and how they are spread. Statistician Brendan Rocks published the formula for its variance in The American Statistician in 2016. With P the share of promoters and D the share of detractors:
- variance per respondent: σ = P + D – (P – D)²;
- margin at 95% (Wald approximation): 1.96 × √(σ / n), for n responses.
For the same score, the margin varies: 100% passives, or half promoters and half detractors, both give an NPS of 0, with a variance of 0 and of 1 respectively. Elev8 Lab calculation for the example (P = 0.5, D = 0.2, σ = 0.61):
| Responses | Margin of error at 95% |
|---|---|
| 20 | ± 34 points |
| 30 | ± 28 points |
| 50 | ± 22 points |
| 100 | ± 15 points |
| 200 | ± 11 points |
| 400 | ± 8 points |
| 1,000 | ± 5 points |
± 15 points
With 100 responses, 50% promoters and 20% detractors, an NPS of 30 has a margin of error of about ± 15 points at 95%.
Elev8 Lab calculation of October 19, 2026, Rocks formula (2016)
Between two waves of the same size, the gap must exceed about 22 points with 100 responses per wave, 15 points with 200 and 11 points with 400. So a 10-point rise proves nothing, even with 400 responses per wave: this is the logic of an A/B test in conversion rate optimization.
Warning
The Wald approximation is mostly reliable above 30 responses; below that, variants (adjusted Wald, iterative score test) do better, according to Rocks.
What is a good NPS?
There is no universal threshold for a good NPS: a positive score only means there are more promoters than detractors. Bain publishes no scale, and the threshold grids found online cite neither a source nor a method.
The only median published by the inventor dates from 2003: 16, across more than 400 companies in 28 industries. It is based on some 130,000 responses collected by Satmetrix since early 2001, from public lists. It is a historical reference point, not a norm. Comparing yourself with your competitors requires three conditions:
- same form of NPS: relationship and experience scores “should not be compared directly”, according to Bain;
- same method: for Bain, the double-blind benchmark is “the best way” to know where you stand;
- same market: according to Rob Markey (Bain, 2014), a lower score in one region may come from a weaker cultural tendency to give a 9 or a 10.
Warning
Be wary of a score announced without a number of responses or a collection method: in 2021, Reichheld himself denounced self-reported, unaudited scores. Read an NPS alongside your other performance indicators.
When and how should you run an NPS survey?
The timing of the survey depends on the form of NPS, which Bain sorts into three in “Three Types of Net Promoter Scores”.
| Form | Trigger | Use |
|---|---|---|
| Competitive benchmark | Double-blind survey run by a third party | See where you stand against competitors |
| Relationship | No trigger, often once or twice a year | Track the overall relationship |
| Experience | An action: online purchase, call to customer service | Assess an interaction |
Your own surveys have two biases: unhappy customers respond less, and repeated sends wear people out. For experience NPS, Bain advises targeting only the important interactions and capping requests in the survey software.
Tip
Add an open question (“Why this rating?”). Send the survey once the request is resolved, not while the customer is waiting for a replacement. By email, apply your usual consent and deliverability rules.
NPS surveys and the GDPR
An NPS survey processes personal data. In France, the CNIL framework on commercial activities (the CNIL is the French data protection authority) lists among its purposes “customer relationship follow-up for carrying out satisfaction surveys” (our translation).
The legal basis is legitimate interest or consent, as reasoned and chosen by the data controller. Data is kept for as long as the survey requires, or until the person objects or withdraws consent. This framework is not binding and excludes health, education, banking, insurance and licensed gambling.
Warning
A survey email that contains a commercial offer also becomes a direct marketing message. It then follows the rules for direct marketing in France: prior consent for consumers, a right to object for professionals, an exception for customers about similar products. This content is not legal advice.
What should you do with NPS responses?
NPS responses show what to fix, customer by customer: Bain calls this “closing the loop”.
- Follow up: call back the customers whose response warrants it.
- Code the verbatims: sort them by theme; an AI summary of verbatims is done on anonymized data.
- Link the rating to the customer record: a named relationship NPS can be matched with actual purchases in the CRM, because customers often overstate their self-reported purchases (Markey, Bain, 2014).
- Cross-check with behavior: compare promoters and detractors in web analytics, for example in Google Analytics 4.
- Feed your personas: verbatims from detractors and passives enrich your personas.
A promoter states an intention, not a published review: to ask for customer reviews, follow the rules on online reviews, without sorting customers by their rating.
NPS, CSAT, CES: what are the differences?
NPS measures the intention to recommend the company, CSAT satisfaction with a specific point, CES the effort made by the customer.
| Metric | Question | Scale | What it measures | When to use it |
|---|---|---|---|---|
| NPS (Net Promoter Score) | Likelihood of recommending the company to a friend or colleague | 0 to 10 | Intention to recommend | Overall relationship, benchmark |
| CSAT (Customer Satisfaction Score) | Satisfaction with an interaction or a product | Short, with no single standard | One-off satisfaction | After a purchase or a contact |
| CES (Customer Effort Score) | “How much effort did you personally have to put forth to handle your request?” | 1 (very low effort) to 5 (very high effort) | Customer effort | Customer service |
CES comes from a study by the Corporate Executive Board (Harvard Business Review, July 2010), carried out with more than 75,000 people. In customer service, it predicts loyalty better than CSAT and NPS. The authors still describe NPS as “a powerful gauge at the company level”. Two CSAT scores, for their part, can only be compared with an identical question and scale.
What are the limits of NPS?
NPS measures an intention that behavior follows poorly, and its predictive power divides Bain and independent research. First gap: Kumar, Petersen and Leone followed 9,900 customers of a telecom operator and 6,700 of a financial services firm (HBR, October 2007). In telecom, 81% thought they would recommend the company and 30% did; in financial services, 68% and 33%.
Second gap: a detractor does not necessarily speak badly of the brand. In May 2018, the agency C Space surveyed more than 2,000 consumers in the United States and the United Kingdom about 10 brands (HBR, October 18, 2019). The score classed 16% of respondents as detractors; only 4% had discouraged someone (self-reported study). Third gap: former customers and non-customers, who voice most negative opinions, are not surveyed (East, Romaniuk, Lomax, 2011).
Warning
The link between NPS and growth is still debated: the studies below differ by industry and by country.
- Bain (NPS Prism method): NPS would explain roughly 20% to 60% of the variation in organic growth between competitors (study by the inventor, non-public data).
- Keiningham et al. (2007): 21 companies, more than 15,500 interviews in Norway; the “clear superiority” of NPS is not replicated.
- van Doorn et al. (2013): on Dutch data, NPS, satisfaction and loyalty are all equally poor at predicting future growth.
- Baehre et al. (2022): 7 sportswear brands in the United States over 5 years; only an NPS measured on all potential customers predicts sales growth.
- Dawes (2022): airlines, supermarkets and insurers; NPS does not indicate future revenue growth.
One last limit, pointed out by the inventor. In “Net Promoter 3.0” (HBR, November 2021), Reichheld, Darci Darnell and Maureen Burns write that firms “corrupted” NPS by turning it into a target. They cite pleading (“I’ll lose my job if you don’t rate me a 10”), gifts in exchange for a 10, and surveys never sent to customers whose claim was denied.
Complementing NPS: the earned growth rate
The earned growth rate is the accounting-based metric proposed in 2021 by Reichheld and his coauthors to complement NPS: NRR + ENC – 100%.
- NRR (net revenue retention): this year’s revenue from last year’s customers, divided by last year’s revenue;
- ENC: share of spending by new customers who came through a referral.
10%
With an NRR of 85% and an ENC of 25%, the earned growth rate is 85% + 25% – 100% = 10%.
Example from Reichheld, Darnell and Burns, “Net Promoter 3.0”, Harvard Business Review, November 2021
ENC requires asking every new customer why they chose you.
Frequently asked questions
What is a good NPS rating?
There is no universal threshold: a positive score only means there are more promoters than detractors. The only reference point published by the inventor dates from 2003, a median of 16 across more than 400 companies. Compare yourself with competitors measured the same way, ideally double-blind, and with yourself over time.
What is the difference between CSAT and NPS?
CSAT measures satisfaction with a specific interaction or product; its question and scale vary from one tool to another. NPS measures the intention to recommend the company, on a fixed 0 to 10 scale and with a single formula. CSAT is asked after a purchase or a contact; relationship NPS is often asked once or twice a year.
What are the 3 metrics for measuring customer satisfaction?
The three metrics most used in surveys are CSAT, NPS and CES. CSAT asks customers whether they are satisfied with an interaction or a product. NPS asks them, from 0 to 10, whether they would recommend the company. CES asks them, from 1 to 5, how much effort they had to make to handle their request. They complement each other.
Is NPS expressed as a percentage?
No. NPS is the difference between two percentages: a number from -100 (all detractors) to +100 (all promoters), written without a % sign. The use of a percentage goes back to the 2003 article, where Reichheld wrote “16%”. Statistician Brendan Rocks noted in 2016 that the score is still “occasionally accompanied by a percentage sign” for presentation.
Should you write “NPS score” or just “NPS”?
Bain asks people not to write “NPS Score”, which could be read as “Net Promoter Score Score”. In English, write “NPS” or “Net Promoter Score”. In French, “score NPS” is common and understood, but “NPS” is enough: “notre NPS est de 30” (our NPS is 30). On a website, add the symbol at the first occurrence and Bain’s attribution notice.
How many responses do you need for a reliable NPS?
There is no official minimum: the margin of error depends on the number of responses and how they are spread. For a score of 30 (50% promoters, 20% detractors), it is about ± 22 points with 50 responses, ± 15 with 100 and ± 8 with 400 (Elev8 Lab calculation, Rocks formula, 2016). Between two waves of 100 responses, only a gap of more than 22 points is significant.
What is eNPS?
eNPS (Employee NPS) applies the Net Promoter Score method to employees: “How likely would you recommend this company as a place to work?” Responses stay confidential, so there is no individual follow-up. According to Bain, the most common frequency is monthly, and employee scores can be much lower than customer scores. Bain calls eNPS “an emerging science”.
Sources
- Bain & Company, Measuring Your Net Promoter Score, undated page. Accessed on October 19, 2026.
- Frederick F. Reichheld, Harvard Business Review, The One Number You Need to Grow, December 2003. Accessed on October 19, 2026.
- Bain & Company, About the Net Promoter System, undated page. Accessed on October 19, 2026.
- Bain & Company, Trademarks and Licensing, undated page. Accessed on October 19, 2026.
- Bain & Company, Three Types of Net Promoter Scores, undated page. Accessed on October 19, 2026.
- Rob Markey, Bain & Company, The benefits of a competitive benchmark Net Promoter Score, February 18, 2014. Accessed on October 19, 2026.
- Bain & Company, Employee NPS, undated page. Accessed on October 19, 2026.
- Brendan Rocks, The American Statistician, Interval Estimation for the “Net Promoter Score”, vol. 70, no. 4, pp. 365-372, October 2016 (published version not consulted, access denied). Accessed on October 19, 2026.
- Brendan Rocks, arXiv, Interval Estimation for the “Net Promoter Score” (preprint 1601.07235), January 27, 2016. Accessed on October 19, 2026.
- Bain & Company, How Net Promoter Score Relates to Growth, undated page. Accessed on October 19, 2026.
- Fred Reichheld, Darci Darnell, Maureen Burns, Harvard Business Review, Net Promoter 3.0, November 2021. Accessed on October 19, 2026.
- V. Kumar, J. Andrew Petersen, Robert P. Leone, Harvard Business Review, How Valuable Is Word of Mouth?, October 2007. Accessed on October 19, 2026.
- Christina Stahlkopf (C Space), Harvard Business Review, Where Net Promoter Score Goes Wrong, October 18, 2019. Accessed on October 19, 2026.
- Timothy L. Keiningham, Bruce Cooil, Tor Wallin Andreassen, Lerzan Aksoy, Journal of Marketing, A Longitudinal Examination of Net Promoter and Firm Revenue Growth, vol. 71, no. 3, pp. 39-51, July 2007 (abstract). Accessed on October 19, 2026.
- Robert East, Jenni Romaniuk, Wendy Lomax, International Journal of Market Research, The NPS and the ACSI: A Critique and An Alternative Metric, vol. 53, no. 3, pp. 327-346, May 2011 (abstract). Accessed on October 19, 2026.
- Jenny van Doorn, Peter S.H. Leeflang, Marleen Tijs, International Journal of Research in Marketing, Satisfaction as a predictor of future performance: A replication, vol. 30, no. 3, pp. 314-318, September 2013 (abstract). Accessed on October 19, 2026.
- Sven Baehre, Michele O’Dwyer, Lisa O’Malley, Nick Lee, Journal of the Academy of Marketing Science, The use of Net Promoter Score (NPS) to predict sales growth: insights from an empirical investigation, vol. 50, no. 1, pp. 67-84, published online on July 5, 2021 (abstract). Accessed on October 19, 2026.
- John G. Dawes, Australasian Marketing Journal, Net Promoter and Revenue Growth: An Examination Across Three Industries, vol. 32, no. 1, pp. 4-18, published online on November 14, 2022 (abstract). Accessed on October 19, 2026.
- Matthew Dixon, Karen Freeman, Nick Toman (Corporate Executive Board), Harvard Business Review, Stop Trying to Delight Your Customers, July 2010. Accessed on October 19, 2026.
- CNIL, Framework for the processing of personal data carried out for the management of commercial activities (in French), PDF dated March 18, 2022. Accessed on October 19, 2026.
- CNIL, Questions and answers on the frameworks for the management of commercial activities and of unpaid debts in a commercial transaction (in French), January 28, 2022. Accessed on October 19, 2026.
- CNIL, Direct marketing by email (in French), undated page. Accessed on October 19, 2026.
Attribution notice requested by Bain & Company: “Net Promoter®, NPS®, NPS Prism®, and the NPS-related emoticons are registered trademarks of Bain & Company, Inc., NICE Systems, Inc., and Fred Reichheld. Net Promoter Score℠ and Net Promoter System℠ are service marks of Bain & Company, Inc., NICE Systems, Inc., and Fred Reichheld.”
Digital marketing, SEO and GEO consultant
Article checked and updated by the author. Sources consulted on the date shown.
Cite this article
Clair, B. (2026, October 24). Net Promoter Score: how do you calculate and read it?. Elev8 Lab. https://elev8-lab.fr/en/customer-loyalty/net-promoter-score/