Net Promoter Score (NPS) Calculator

Background.

Last quarter your NPS was 44. This quarter 100 customers answered the survey: 62 gave a 9 or 10, 25 gave a 7 or 8 and 13 gave 6 or less, for a score of 49. Is that five-point rise real? With 100 responses, the 95% confidence interval around 49 runs from 33.6949 to 61.4507, so last quarter's 44 sits comfortably inside it and the survey cannot tell the two quarters apart. Four times as many responses in the same proportions narrows the range to 41.6519 to 55.6185; ten thousand narrows it to 47.5859 to 50.3847. That interval is printed under every score because it is the difference between a trend and noise.

The score itself comes from one question, "How likely are you to recommend us to a friend or colleague?", answered on a scale from 0 to 10. Bain & Company's definition sorts the answers into three groups, promoters at 9 or 10, passives at 7 or 8 and detractors at 0 to 6, and the Net Promoter Score is the percentage of promoters minus the percentage of detractors. It can run from −100 to +100 and is normally quoted as a whole number.

You can enter the three group counts, or paste the individual ratings and let the calculator sort them. Ratings must be whole numbers from 0 to 10; half points and values such as 11 are rejected with an error. A one-to-five star scale is a trap: those answers are all valid numbers between 0 and 10, so they are accepted and every one of them lands among the detractors. When every pasted rating is 5 or lower, the summary line under the score carries a caution that the data look like a 1–5 scale.

Treat the score as a consistent way to track loyalty over time rather than a forecast. Fred Reichheld introduced it in the Harvard Business Review in 2003 as the one number a company needs to grow, but a later study of Norwegian firms could not reproduce the claim that it beats other loyalty measures.

What is net promoter score calculator?

The Net Promoter Score summarises answers to the recommendation question in one figure: the share of respondents who are promoters minus the share who are detractors, times 100. Passives do not appear in the subtraction, but they are counted in the total that both shares are taken from. An equivalent way to see it, used by the statistician Brendan Rocks, is to score each promoter +1, each passive 0 and each detractor −1 and take the average; NPS is that average times 100. Because it is a difference between two percentages, the score is not itself a percentage and is usually written without a % sign: +49, not 49%. Two very different customer bases can share a score, so the promoter, passive and detractor shares are shown with it. Net Promoter, NPS and Net Promoter Score are trademarks or service marks of Bain & Company, NICE Systems and Fred Reichheld.

How to use this calculator.

1. 01Pick the input mode: group counts if your survey tool has already sorted respondents into promoters, passives and detractors, raw ratings if you have each person's 0–10 answer.
2. 02For group counts, enter all three as whole numbers of people. Include the passives; leaving them out shrinks the total and inflates the score.
3. 03For raw ratings, paste the answers separated by commas, spaces or new lines. Make sure the survey used the 0–10 scale, because answers from a 1–5 scale will be accepted and scored as detractors; if every rating is 5 or lower, the summary line warns that the data look like 1–5.
4. 04Read the score with its 95% interval. When this period's interval contains last period's score, the survey has not shown a change.
5. 05Look at the three shares as well as the score, since an even split between fans and critics and a base of lukewarm customers can produce the same number.
6. 06Keep the question, the scale, who gets surveyed and when the same from period to period; the interval cannot account for a change in method.

The formula.

With P promoters, D detractors and N respondents in total, passives included, NPS = (P ÷ N × 100) − (D ÷ N × 100). Writing p₊ = P ÷ N and p₋ = D ÷ N, the score on a −1 to +1 scale is p₊ − p₋, and the calculator reports it times 100.

The confidence interval treats each response as a draw of +1, 0 or −1. The variance of a single draw is p₊ + p₋ − (p₊ − p₋)², and dividing by the number of responses gives the variance of the score. A plain Wald interval would be the score plus or minus 1.96 times the square root of that. Rocks, writing in The American Statistician in 2016, found the Wald interval performs poorly for NPS and recommended an adjusted version he labelled AW(3,T), which is the one used here. It adds 0.75 to the promoter count, 0.75 to the detractor count and 1.5 to the passive count, so the adjusted total is N + 3, then builds the Wald interval from the adjusted shares. The 1.96 is the point on the normal distribution that leaves 2.5% in each tail, giving 95% coverage.

The adjusted score works out to (P − D) ÷ (N + 3), the same net count spread over three extra imaginary respondents, so it sits slightly closer to zero than your score. That is why the interval is not symmetric about your score; the effect is large for small samples and vanishes for big ones. It also stops the interval collapsing to zero width when every answer falls in one group. Because the adjusted interval can stretch past −100 or +100, its ends are cut off at those limits.

The width depends on two things: the number of responses, through the square root, and how spread out the answers are. A base split evenly between promoters and detractors has the largest variance, and a base where everyone falls in the same group the smallest, so two surveys with the same score and size can have very different intervals.

The primary result is the unrounded score. A second result gives it as a whole number, with exact halves rounded away from zero, so +0.5 becomes +1 and −0.5 becomes −1. Bain's definition does not specify a rounding rule, so that is a reporting convention; what matters is applying the same one every period.

A worked example.

Frequently asked questions.

References& sources.

1. [1]Bain & Company, "Measuring Your Net Promoter Score℠" (netpromotersystem.com, current edition, retrieved 2026-07-29). The issuing body's own definition: "To calculate NPS, start with the ultimate question, 'How likely are you to recommend us to a friend or colleague?' and score the answers on a zero-to-ten scale. Your Net Promoter Score is simply the percentage of customers who are promoters (those who scored 9 or 10) minus the percentage who are detractors (those who scored 0 to 6)." Also the source of the band labels Promoters (9 to 10), Passives (7 or 8), Detractors (0 to 6), the rationale for the 11-point scale, the statement that passives' "repurchase and referral rates are as much as 50% lower than those of promoters", and the trademark notice. Publicly accessible.
2. [2]Reichheld, F. F. (2003). "The One Number You Need to Grow." Harvard Business Review, December 2003 issue. The original publication of the metric and of the growth claim attached to it. Author, title and issue date verified on the publisher's page 2026-07-29; the full article text is paywalled, so this is cited as the origin of the metric rather than for any specific figure.
3. [3]Rocks, B. (2016). "Interval Estimation for the 'Net Promoter Score'." The American Statistician, 70(4), 365–372, DOI 10.1080/00031305.2016.1158124; open preprint arXiv:1601.07235, retrieved and text-extracted 2026-07-29. Supplies the variance of the score, σ = p_pro + p_det − (p_pro − p_det)², the finding that "The Wald and Goodman tests perform poorly; their use should be avoided", and the recommended method implemented here: "The best performing Adjusted Wald is AW(3,T) which can be used by adding 3/4 to the counts of both Promoters and Detractors, and 3/2 to the count of Passives, before construction of a Wald interval." Journal version paywalled; preprint open.
4. [4]Keiningham, T. L., Cooil, B., Andreassen, T. W. & Aksoy, L. (2007). "A Longitudinal Examination of Net Promoter and Firm Revenue Growth." Journal of Marketing, 71(3), 39–51, DOI 10.1509/jmkg.71.3.039 (abstract retrieved 2026-07-29). Consulted as an independent authority on the metric's predictive claim rather than its arithmetic. Using 21 firms and more than 15,500 interviews from the Norwegian Customer Satisfaction Barometer, "the research fails to replicate his assertions regarding the 'clear superiority' of Net Promoter compared with other measures in those industries." Abstract public; full text paywalled.
5. [5]Agresti, A. & Coull, B. A. (1998). "Approximate Is Better Than 'Exact' for Interval Estimation of Binomial Proportions." The American Statistician, 52(2), 119–126, DOI 10.1080/00031305.1998.10480550. The origin of the adjusted-Wald construction that Rocks (2016) adapts to the trinomial Net Promoter case; cited here because the AW(3,T) interval on this page is a direct descendant of it. Bibliographic reference — the publisher's copy is paywalled and was not fetched; the method as implemented is taken verbatim from Rocks §2.1.3 and §4.