Net Promoter Score (NPS) Calculator
Your NPS with the margin of error it comes with, so a five-point move on two hundred responses is reported as noise rather than as progress.
Promoters, 9 to 10 312 (64.5%)
Passives, 7 to 8 118 (24.4%)
Detractors, 0 to 6 54 (11.2%)
Responses 484
Net Promoter Score 53
95% margin of error ± 6.1 points
Likely range 47.2 to 59.4
Target score 60
Detractors to convert 17
The score is 53: 64.5% promoters less 11.2% detractors. The 118 passives
count in the total and nowhere else, which is why a company can move a
third of its answers and not move its score at all.
On 484 responses the 95 percent interval is 47.2 to 59.4. Any change
smaller than 6.1 points between two surveys this size is inside the
noise, so treat it as flat rather than as a trend. To halve that margin
you need roughly four times the responses.
A score of 53 is high. Comparisons to a published industry average are
weaker than they look: the number moves with the wording of the
question, when in the relationship you ask, and whether you survey
everyone or only the people who answer support tickets.
Reaching 60 on this sample size means moving 17 detractors all the way
to promoters. Each one is worth two shares rather than one, because it
leaves the bottom group and joins the top.
Output is valid and updates as you type.
Fix the highlighted fields to update the output.
The score is subtraction: the share of promoters less the share of detractors, with the passives counted in the total and nowhere else. Every calculator does that part.
What most of them leave out is that the number is an estimate from a sample, and a noisy one. NPS throws away the middle of the scale, which makes its variance larger than a plain percentage’s, so on a few hundred responses the margin of error is routinely six points or more. This reports the interval next to the score, because a five-point quarter-on-quarter “improvement” on 200 responses is noise, and a team that reports it as progress will spend a quarter chasing it.
How to use
- Count the responses in each group: 9 and 10 are promoters, 7 and 8 are passives, 0 through 6 are detractors.
- Put the three counts in.
- Optionally set a target score to see how many detractors would have to become promoters to reach it on this sample.
Example
312 promoters, 118 passives, 54 detractors:
Promoters, 9 to 10 312 (64.5%)
Passives, 7 to 8 118 (24.4%)
Detractors, 0 to 6 54 (11.2%)
Responses 484
Net Promoter Score 53
95% margin of error ± 6.1 points
Likely range 47.2 to 59.4
Target score 60
Detractors to convert 17The score is 53 and the honest statement is “somewhere between 47 and 59”. If last quarter was 49 on a similar sample, nothing measurable happened.
Pitfalls
A score without a sample size is not a result. Fifteen responses can produce an NPS of 60 with an interval so wide it includes zero. Ask for n before you react to the number, and treat any change smaller than the margin of error as flat.
Halving the error takes four times the responses. The interval shrinks with the square root of the sample, so going from 100 responses to 400 halves it and going from 400 to 500 barely moves it. That is the whole budget conversation about survey size in one sentence.
Passives are not neutral in the arithmetic, they are in the denominator. Moving a third of your answers from 7 to 8 changes nothing at all. The score only reads the two ends, which is a real criticism of the metric and worth knowing before you set a bonus on it.
Benchmarks are weaker than they look. The score moves with the wording of the question, when in the relationship you ask, and who you ask. A survey sent to everyone and a survey sent after a support ticket measure different things, and neither is comparable to an industry average collected a third way.
Who answers is usually the bigger problem. Response rates are low and the people who answer are not a random sample: they are the delighted and the furious. Nothing in this calculator can fix selection bias, and it is larger than the sampling error it does report.
Rounding, and the 6. The score is conventionally a whole number. And the detractor group runs to 6 inclusive: a 6 out of 10 is a detractor, which surprises people the first time they see it and is the standard definition.
Compatibility
Arithmetic in the browser: nothing is uploaded and nothing is stored, which matters here because survey data often is not yours to send anywhere.
The margin of error uses the normal approximation with the variance of the promoter-minus-detractor difference, which is p + d – (p – d)^2 once the passives are collapsed into the total. That is the standard multinomial result. It is a good approximation for a few hundred responses and it gets shaky below about thirty, or when almost every answer is in one group: at 100 percent promoters the variance goes to zero and the interval with it, which is arithmetically true and practically misleading.
The interval assumes the respondents are a random sample of your customers. They are not. Treat it as the floor on your uncertainty, not the whole of it.