Net Promoter Score calculator
Promoters minus detractors, with the margin of error most survey tools leave out.
How is the Net Promoter Score calculated?
The survey asks one question: how likely are you to recommend us to a friend or colleague, on a scale from 0 to 10? Answers of 9 or 10 are promoters, 7 or 8 are passives and 0 to 6 are detractors. The score is the percentage of promoters minus the percentage of detractors, so it can run from −100 (everyone is a detractor) to +100 (everyone is a promoter), and passives only count in the base. The idea was popularized by Frederick Reichheld in the December 2003 Harvard Business Review article “The One Number You Need to Grow”. Example: of 100 responses, 50 promoters, 30 passives and 20 detractors give 50% − 20% = +30. NPS is written as a plain number, not a percent.
Worked example: 65 promoters, 68 passives, 79 detractors
That is 212 responses, with 30.7% promoters and 37.3% detractors, so NPS = −6.6. For the margin of error we need the variance of the score. Each response counts as +1 (promoter), 0 (passive) or −1 (detractor), which makes the variance p + d − (p − d)² = 0.67, and the standard error is √(variance ÷ n) = 5.64 points. At 95% confidence the margin of error is 1.96 × 5.64 = ±11.1 points, so the interval runs from −17.7 to +4.5. The interval includes zero, which means this survey cannot tell whether the true NPS is negative or positive.
How big a sample do I need?
NPS needs more responses than most people expect, because it combines two proportions and the variance is large. When half of respondents are promoters and a fifth are detractors (NPS +30), the variance is 0.61 and a ±5 point margin at 95% confidence takes 938 responses. At only 20 responses with 8 promoters, 6 passives and 6 detractors the score is +10.0, but the interval runs from −26.4 to +46.4, which is almost the entire useful range. The calculator shows the responses needed for ±5 and ±10 points at your own promoter and detractor shares. If you cannot reach those sizes, report the interval, not just the number, and avoid reading anything into a 3-point move.
Is a change in NPS from one quarter to the next real?
Not necessarily. Each quarter’s score has its own margin of error, and the difference between two independent scores is uncertain by more than either one alone: the standard error of a difference is √(SE₁² + SE₂²). With about 200 responses each, a quarter-to-quarter change has to be on the order of 15 points before it is clearly outside noise. Compare the intervals before you celebrate or panic. MeasuringU describes a significance test for the difference between two NPS values built on the same variance formula.
Why is the interval clipped at −100 and +100?
The normal approximation does not know the score cannot exceed 100 in absolute value. With few responses and a score near the limits, the computed interval can run past them; the calculator clips it there and says so. If every response lands in one group, the variance formula returns zero and the interval collapses to a single point, which is a limitation of the formula, not proof of certainty. MeasuringU recommends an adjusted-Wald correction (adding 3 pseudo-responses) for small samples; this page uses the plain formula and warns you when n is small.
Is NPS a good predictor of growth?
Reichheld’s 2003 article argued that the recommend question was the single best predictor of growth, and NPS became a widely used management metric. Critics have noted that the original evidence compared past growth rates with scores, not future growth, and that one number can hide what customers actually want fixed. Treat NPS as a tracking signal and read the open-text reasons alongside it. A comparison across industries is also hazardous, since average NPS differs a lot between sectors and survey methods.
Common mistakes
Counting 7s and 8s as promoters is the usual slip: passives belong in the base but not in the numerator. Another is quoting NPS as a percent; it is a number from −100 to +100. Averaging the 0–10 ratings is a different statistic and does not give an NPS. Finally, do not compare scores from different survey channels or wording without noting it, because the same customers answer differently on a website pop-up and an e-mail survey.
For general information only. Sources: F. F. Reichheld, “The One Number You Need to Grow,” Harvard Business Review, December 2003, pp. 46–54 (citation confirmed from library records; the article itself was not opened); MeasuringU, “How to Statistically Compare Two Net Promoter Scores” (variance of NPS and its adjusted-Wald variant). Related: margin of error calculator, sample size calculator and confidence interval calculator. How we check our calculators.