Back
NPS Calculator: Work Out Your Net Promoter Score
Matthieu Saussaye
The Short Answer
Enter how many people scored 9-10, 7-8 and 0-6. The calculator returns NPS = %Promoters - %Detractors, on a scale from -100 to +100.
Passives stay in the total, but add nothing to the score.
The result is a number, not a percentage.
Only compare it to your own past scores.
Worked Examples
Five scenarios, each with 200 responses, so you can sanity-check what the calculator gives you. Notice that examples C, D and E all land near zero for completely different reasons.
Scenario | Promoters (9-10) | Passives (7-8) | Detractors (0-6) | NPS |
|---|---|---|---|---|
A. Strong advocacy | 140 (70%) | 40 (20%) | 20 (10%) | +60 |
B. Healthy | 90 (45%) | 70 (35%) | 40 (20%) | +25 |
C. Polarized | 80 (40%) | 40 (20%) | 80 (40%) | 0 |
D. Indifferent | 0 (0%) | 200 (100%) | 0 (0%) | 0 |
E. Slipping | 50 (25%) | 60 (30%) | 90 (45%) | -20 |
Scenarios C and D produce the same headline number and demand opposite responses. In C you have a product that some people love and others actively dislike, which usually means a segment problem. In D nobody feels anything, which is a differentiation problem. This is the clearest argument for always reporting the three-way split next to the score.
What to Enter
The calculator takes three counts, not percentages.
Promoters: the number of people who answered 9 or 10.
Passives: the number who answered 7 or 8.
Detractors: the number who answered 0, 1, 2, 3, 4, 5 or 6.
Count only people who actually answered the recommend question. Someone who opened your survey and skipped the question is not a detractor, and someone who never opened it is not a respondent at all. If you export from a survey tool, filter out blanks before you count. If your scale runs 1-10 rather than 0-10, note that in your documentation, because the buckets are defined for a 0-10 scale and your results will not be directly comparable to anyone using the standard wording.
How to Read the Result
Three things to look at, in this order.
1. The sign
Positive means promoters outnumber detractors. Negative means the reverse. Zero means they cancel, which as the table above shows can mean two very different things.
2. The distribution behind it
Look at the three percentages, not just the net. A score of +25 built from 45% promoters is a different business from a +25 built from 30% promoters and only 5% detractors. The first has passion and a problem. The second has broad, mild approval.
3. The change from last wave
The absolute value is close to meaningless without a comparison. What matters is the direction against your own previous measurement, taken on the same population with the same question wording, the same channel and the same trigger. Change any of those and you have broken the comparison, whatever the numbers say.
What you cannot read from it
The score does not tell you why anything moved. That comes from the open follow-up question, and only once those responses are coded into themes you can count. A full breakdown of what the metric does and does not capture is in our guide to Net Promoter Score.
Common Calculation Mistakes
Counting passives in the numerator
The most frequent error. Passives are excluded from the top of the calculation but included in the base. If you have 45% promoters, 35% passives and 20% detractors, NPS is 45 - 20 = +25. It is not 45 + 35 - 20, and it is not calculated by dropping passives and recomputing promoters and detractors as shares of the remaining 65%. That second version would give you +38 for the same data, a 13-point error in your favor, which is exactly why it keeps happening.
Reporting it as a percentage
NPS is a difference between two percentages, which makes it a number on a -100 to +100 index, not a percentage itself. "NPS of 42%" is a giveaway that the number has passed through a slide deck without anyone checking it.
Sample size too small
Because NPS combines two proportions, its margin of error is wider than the margin on a single proportion at the same base size. At small bases, a couple of respondents flipping between 8 and 9 can swing the reported score by double digits with no real change underneath. This bites hardest when you cut by segment: the overall number is stable, the per-country or per-product-line numbers bounce around, and someone builds a narrative out of the bouncing. Decide a minimum base for reporting, publish base sizes on every chart, and suppress cuts that fall below it.
Comparing across industries
Categories differ in how recommendable they are at all. Some products are talked about at dinner; others are infrastructure people only notice when they break. A gap between a software company and a utility says more about the category than about either company's performance. Compare like with like, or compare only against yourself.
Comparing across countries without a caveat
Rating-scale behavior differs by country and language. Cross-cultural survey research has consistently found that willingness to use the top and bottom of a scale is not constant across populations. Since NPS only rewards 9s and 10s, markets where respondents avoid scale extremes will show structurally lower scores for the same underlying sentiment. Track each market against its own trend rather than ranking markets against each other.
Mixing transactional and relational responses in one number
Post-support-ticket responses and scheduled relationship surveys come from different samples. Support surveys over-represent people who just had a problem. Pooling them produces a number that belongs to neither population. Keep two lines on the chart.
Changing the send logic mid-track
Switching from email to in-app, moving the survey earlier in the journey, or excluding a cohort will move the score on its own. If you must change the method, run both in parallel for one wave so you can measure the size of the shift, and annotate the break on the chart.
After the Number: What to Actually Do
A calculator gives you a figure. The figure does not improve anything by itself. Three steps turn it into work.
Add the "why" question. Ask each group a different follow-up. Detractors: what went wrong and what should have happened instead. Passives: the one change that would have made it a 9. Promoters: what they actually tell other people about you.
Code the answers into themes. Until open responses are quantified, they lose every argument against a chart. You want to be able to say a theme accounts for a specific share of detractor comments, and to click through to the raw verbatims behind that share.
Route detractors to a human. Close-the-loop contact within a working day or two is where most of the retention effect of an NPS program comes from, not from the measurement.
SmartInterview covers steps one and two directly: an AI follow-up probes the reason behind the score in the respondent's own language, by voice or text, and the open responses are coded automatically into themes you can count and audit against the original text.
Related Reading
Frequently Asked Questions
How do you calculate NPS manually?
Divide your promoter count (9s and 10s) by the total number of people who answered the question, and do the same for detractors (0 to 6). Convert both to percentages, then subtract the detractor percentage from the promoter percentage. With 90 promoters, 70 passives and 40 detractors out of 200 responses, that is 45% - 20% = +25.
Do passives count toward NPS?
Passives are counted in the total number of respondents but are neither added nor subtracted in the formula. Leaving them out of the denominator is the single most common way people accidentally inflate their score.
Can NPS be negative?
Yes. Any time detractors outnumber promoters, the score is negative. The theoretical minimum is -100, which would require every respondent to score 6 or below.
What is the highest possible NPS?
+100, reached only if every single respondent gives a 9 or a 10. In practice a score near the top of the range is far more likely to indicate a biased sample than universal enthusiasm, so check who was surveyed before celebrating.
How many responses do I need before the score means anything?
More than the number most dashboards will happily display. NPS is a difference of two proportions, so it is noisier than a single proportion at the same base. Set a minimum base per reporting cut, show base sizes on every chart, and treat movement in small segments as noise until it repeats across waves.
Is NPS out of 10 or out of 100?
The question is asked on a 0-10 scale. The score derived from it runs from -100 to +100. The two scales are often confused, which is why a reported "NPS of 8" almost always means someone has published an average rating rather than a Net Promoter Score.
Why did my NPS change when nothing else did?
Usually the survey changed, not the customers. Check whether the send channel, the timing in the journey, the audience filter or the question wording moved between waves. Also check the base size, since a small sample can shift by double digits on a handful of responses.
