NPS Calculator
The subtraction in an NPS calculator takes one line. The part everyone skips is that NPS on 200 responses carries about ten points of margin, so most quarter-on-quarter movement is noise.
Net Promoter Score
- Net Promoter Score 53.5% promoters minus 18.8% detractors, across 340 responses (good). The score is a number between -100 and 100 and is not a percentage, which is why it is quoted without a sign.
- 35
- 95% confidence interval Plus or minus 8.3 points. This is the figure almost no NPS tool reports, and it is usually wider than the movement being discussed. On 340 responses you cannot distinguish a 35 from a 42.
- 26 to 43
- Change since last period NOT a real change. The two intervals overlap: a gap of 6 points is inside the 13.5 points that sampling alone explains at these response counts. Reporting it as movement is reporting noise.
- +6 points
- Promoters Scored 9 or 10. Only two of the eleven options count as a promoter, which is why the scale is harsher than it looks.
- 182 (53.5%)
- Passives Scored 7 or 8. They are in the denominator and in neither term of the subtraction, so they drag the score toward zero without being unhappy. Converting all 94 of them to promoters would take the score to 62.
- 94 (27.6%)
- Detractors Scored 0 to 6, seven of the eleven options. A 6 counts the same as a 0, which is the most criticized property of the metric.
- 64 (18.8%)
- Responses needed for +/- 5 points At this mix of promoters and detractors. You have 340. Precision improves with the square root of the sample, so halving the margin needs four times the responses.
- 927
- Response rate 340 of 1,400. Below about 30% the non-response is the dominant uncertainty and it is not random: people with strong opinions answer surveys. The confidence interval above describes sampling error only and cannot see this at all.
- 24.3%
| Responses | Margin of error | A score of 35 means |
|---|---|---|
| 50 | +/- 21.5 | 13 to 57 |
| 100 | +/- 15.2 | 20 to 50 |
| 250 | +/- 9.6 | 25 to 45 |
| 500 | +/- 6.8 | 28 to 42 |
| 1,000 | +/- 4.8 | 30 to 40 |
| 5,000 | +/- 2.2 | 33 to 37 |
- NPS is a score from -100 to 100, not a percentage. An NPS of 35 does not mean 35% of anything.
- The confidence interval covers sampling error only. It says nothing about who chose not to reply, which in a survey with a low response rate is the larger problem by far.
How the NPS Calculator works
Net Promoter Score asks one question (how likely are you to recommend us, from 0 to 10) and turns the answers into a single number. Scores of 9 and 10 are promoters, 7 and 8 are passives, and 0 through 6 are detractors. NPS is the percentage of promoters minus the percentage of detractors, which puts it between -100 and 100.
With 182 promoters, 94 passives and 64 detractors out of 340 responses, that is 53.5% minus 18.8%, or 35.
Three things about the arithmetic are worth noticing because they are deliberate and widely misread. It is a score and not a percentage: an NPS of 35 does not mean 35% of anything. Passives appear in the denominator and in neither term of the subtraction, so they pull the score toward zero without ever moving it directly. And the bands are lopsided: seven of the eleven options make someone a detractor and only two make them a promoter, so a 6 out of 10 counts exactly as badly as a 0.
The part that matters more than any of it, and that almost no NPS tool reports, is that this is a sample statistic and it has a margin of error. NPS is a trinomial mean, so its standard error is the square root of (promoter share + detractor share - NPS squared, as a fraction) over the number of responses. On 340 responses at this mix the 95% interval is about plus or minus 8 points. On 100 responses it is about 15.
That has an uncomfortable consequence for how NPS is usually used. A quarterly score that goes from 29 to 35 on a few hundred responses has not moved: the intervals overlap heavily, and the difference between two independent samples needs to clear roughly the combined margin before it means anything. A great deal of management attention is spent on NPS movements that are arithmetic noise, and the tooling encourages it by reporting a single confident-looking integer.
Precision improves with the square root of the sample, so halving the margin requires four times the responses. Getting to plus or minus 5 points usually takes somewhere between 800 and 1,000 responses, and to plus or minus 2 points, several thousand. For most companies the honest conclusion is that NPS is a useful annual direction-of-travel figure and a poor quarterly KPI.
And the confidence interval is the smaller of the two uncertainties. It describes sampling error: the luck of which customers happened to be asked. It says nothing about who chose to answer, and survey response is emphatically not random. People with strong feelings reply, and the indifferent middle does not. At a 20% response rate, the 80% who said nothing are a larger unknown than anything the statistics can quantify, and no amount of extra responses fixes it if the bias is in who responds rather than how many.
Used well, NPS is a tracking metric for one population measured the same way over time, with the interval reported alongside it and the verbatim comments read. Used badly it is an integer on a dashboard that people are compensated against, which produces score-chasing (asking only happy customers, timing the survey after a good interaction) long before it produces a better product.
NPS Calculator FAQ
How do you calculate NPS?
Classify each 0-10 response: 9-10 promoters, 7-8 passives, 0-6 detractors. NPS is the percentage of promoters minus the percentage of detractors, with passives counted in the denominator only. 182 promoters and 64 detractors out of 340 responses gives 53.5% - 18.8% = 35. The result is a score between -100 and 100, not a percentage.
What is a good NPS?
As a rough guide, above 0 means more promoters than detractors, 30+ is good, 50+ is excellent and 70+ is world class. Benchmarks vary enormously by industry and by country (some cultures score lower on eleven-point scales), so comparing your figure with your own previous one is far more informative than comparing it with a published average.
What is the margin of error on NPS?
Larger than most people assume. NPS is a sample statistic whose standard error is the square root of (promoter share + detractor share - NPS as a fraction, squared) divided by the number of responses. On 340 responses that is roughly plus or minus 8 points at 95% confidence; on 100 responses it is about 15. Reaching plus or minus 5 points usually needs 800 to 1,000 responses.
Is a change in NPS from 29 to 35 significant?
On a few hundred responses, almost certainly not. Two independent samples each carry their own margin, and the difference has to clear roughly the combined figure before it is distinguishable from sampling noise. Six points on 300-odd responses is well inside it. Most quarter-over-quarter NPS movement reported in dashboards is noise.
Why do passives not count in NPS?
By design: the metric is meant to contrast enthusiasm with dissatisfaction, and 7s and 8s are neither. They are not ignored, though: they sit in the denominator, so a large passive group drags the score toward zero. Converting passives to promoters is usually the cheapest way to move the number, since they are already not unhappy.