NPS Calculator

Ask "How likely are you to recommend us to a friend or colleague?" on a 0–10 scale. Answers of 9 and 10 are promoters, 7 and 8 are passives, 0 to 6 are detractors.

NPS = % of promoters − % of detractors. Passives count in the total but not in either share, so the score runs from −100 (everyone a detractor) to +100 (everyone a promoter). It is written as a plain number, "NPS 20", not "20%".

A score from a survey is an estimate. The calculator shows the range that fits your answers and, for planning, how many answers a narrower range takes.

200 people answered: 90 promoters, 60 passives and 50 detractors. Promoters are 45.0%, detractors 25.0%, so the NPS is 45 − 25 = 20.

The 95% range is 8.5 to 30.9. A score anywhere in that range fits these 200 answers, so a move from 20 to 25 next month could easily be noise.

With only 36 answers (15 promoters, 13 passives, 8 detractors) the NPS is 19.4 and the 95% range runs from −6.3 to 42.2: the survey cannot even tell whether the score is positive. For ±5 points at the example's mix you'd need about 1,015 answers.

Score each answer +1 (promoter), 0 (passive) or −1 (detractor): the NPS is 100 × their average. The spread of these values gives the range. Because the NPS is a difference of two shares from the same people, the range of a single proportion does not apply.

n = P + Pa + D
NPS = 100 × (P − D) / n
nA = n + 3,  pP = (P + 0.75) / nA,  pD = (D + 0.75) / nA
c = pP − pD,  V = pP + pD − c²
range = 100 × (c ± q × √(V / nA)), within −100 … +100

The simple formula, NPS ± q × standard error, fails where people look: with every answer a promoter it gives a range of zero width, and at small samples with lopsided answers it misses the true score far more often than stated. The calculator adds a few imaginary answers first (three quarters of an answer to promoters and to detractors, three to the total) and then applies the simple formula. The range stays close to the stated level even at 10 answers and at extreme mixes, and it is slightly uneven around the score.

The number of answers for a margin ± m points is q² × V / (m / 100)², where V is the spread of the answers. With your current mix V comes from your data; the worst case, half promoters and half detractors, is V = 1.

For two periods the calculator treats them as independent samples and builds a range for the change from the same adjusted numbers.

How do you calculate NPS?

Count the promoters (9–10), passives (7–8) and detractors (0–6). Divide promoters and detractors by the total number of answers and subtract: NPS = % promoters − % detractors. 90 promoters, 60 passives and 50 detractors give 45% − 25% = 20. The result is a score from −100 to +100, not a percentage.

What is a good NPS?

There is no single threshold. Scores depend on the industry, on who you ask and on when you ask (right after purchase, after a support ticket, once a year). The most useful comparison is with your own earlier surveys run the same way, and only after checking the range: a change inside the noise is not a change.

Why does my NPS have a margin of error? How many responses do I need?

Your answers are a sample of all customers, so another sample of the same size would give a slightly different score. With 200 answers the 95% range is about ±11 points; for ±5 points you need roughly 1,000 to 1,500 answers, depending on the mix. The planner under the result gives the number for your data.

Is a change in NPS from one month to the next significant?

Turn on "Compare with another period" and read the range for the change. If it includes 0, the change could be noise. Two ranges that overlap don't prove there is no change, so compare the change's own range with 0. The comparison assumes the two periods are answered by different people.

Why are 7 and 8 ignored?

They are not ignored: passives count in the total, so they lower both the promoter and the detractor share. They just don't push the score up or down themselves. Adding 50 passives to the example (90 / 110 / 50) moves the NPS from 20 to 16.

Can I average the 0–10 scores instead?

You can, but it is a different metric. The average hides how split the answers are: a mix of 10s and 0s and a column of 5s can have the same mean and very different NPS values. Report the NPS, and keep the average only as a side note if your team is used to it.

Which method does the calculator use for the range?

An adjusted version of the simple formula: it adds three quarters of an answer to promoters and to detractors before computing the range. It stays reliable with small samples and when nearly everyone gives the same answer. The "Method and formulas" section above has the details.

Is my survey data sent anywhere?

No. The calculation runs in your browser. Counts and pasted answers are not stored, sent to a server, put in the address or recorded by analytics.

For a single rate such as a sign-up rate, with its range and the sample size you need: Confidence interval calculator

• Net Promoter Score (NPS)
• confidence interval (CI)
• margin of error

• Confidence interval calculator

  Seeing how precise a single conversion rate is.

• Cohort retention calculator

  Reading a cohort retention table: heatmap and average curve.

• Churn rate and NRR calculator

  Customer churn, revenue churn, GRR and NRR for one period.