How this instrument works
Net Promoter Score comes from a single survey question: how likely are you to recommend this company, product, or service to a friend or colleague, answered on a 0-to-10 scale. Respondents scoring 9 or 10 are promoters, respondents scoring 0 through 6 are detractors, and respondents scoring 7 or 8 are passives. Fred Reichheld introduced the method in a 2003 Harvard Business Review article, arguing that this one differential predicted revenue growth better than the longer satisfaction surveys most companies were running at the time.
The arithmetic is a subtraction, not an average and not a ratio: NPS is the percentage of promoters minus the percentage of detractors, both measured against the full response count. Passives never enter the numerator, but they still shrink the score indirectly, because a larger passive share leaves less room for the promoter and detractor percentages to be large themselves. A product lead or CX manager tracks this figure wave over wave, after a redesign, a pricing change, or a support-process fix, to see whether the shift moved loyalty or just moved feature requests.
The score also hides scale by design. A reading of 50 built from twelve responses and a reading of 50 built from twelve thousand responses look identical on this page, even though the twelve-response version can jump thirty points on the next survey purely from which few people happened to answer. The instrument returns the subtraction correctly either way; judging whether that number means anything yet is a separate question the response count answers, not the formula.
- Enter the share of respondents who scored 9 or 10 on the recommendation question as Promoters, %.
- Enter the share who scored 0 through 6 as Detractors, % — leave the 7-8 passives out of both fields.
- Read Net Promoter Score — the instrument subtracts Detractors, % from Promoters, % for you.
- Recompute after every survey wave using the same question wording and the same 0-10 scale, so one reading compares cleanly against the next.
Worked example — 60% promoters, 10% detractors
Take a product team closing out a quarterly relationship survey of 500 respondents, each answering one question: how likely are they to recommend the product to a friend or colleague, scored 0 to 10. Sixty percent of respondents scored 9 or 10 and ten percent scored 0 through 6; the remaining thirty percent, scoring 7 or 8, are passives and never enter either bucket. Enter 60 as Promoters, % and 10 as Detractors, %, and the instrument subtracts one from the other to return a Net Promoter Score of 50.
A Net Promoter Score of 50 sits well inside the range most benchmarking studies call excellent — the scale itself runs from −100, meaning every respondent is a detractor, to 100, meaning every respondent is a promoter, and many industry surveys treat anything above 50 as rare, world-class territory. What the number does not show on its own is the thirty percent who felt lukewarm enough to sit out of both counts, or how many actual people the sixty and ten figures represent.
Questions
Why does the formula ignore respondents who scored 7 or 8?
The 7-to-8 group, called passives, sits outside the calculation because Reichheld's original research found their future behavior unpredictable — sometimes loyal, sometimes one bad experience from leaving. Leaving them out keeps the score built from the two groups the research found most consistent, but passives still shrink the reading indirectly: a bigger passive share leaves smaller promoter and detractor percentages to subtract, since every respondent still counts toward the survey total.
Is a Net Promoter Score a percentage?
No, even though both inputs are entered as percentages. NPS is the difference between two percentages, a point spread that can land anywhere from −100 to 100, not a share of anything. A score of 50 is usually written and spoken as 'fifty,' never 'fifty percent' — reading it as the share of respondents who are somehow satisfied is the most common misreading of this figure.
What counts as a good Net Promoter Score?
It depends heavily on industry, since baseline loyalty in software differs sharply from baseline loyalty in cable television or airlines. Widely cited benchmarking studies describe positive scores above zero as acceptable, scores above 30 as good, and scores above 50, like the worked example here, as excellent — but a raw score compared against a different industry's benchmark, or against no benchmark at all, tells a team very little on its own.
Does the sample size behind these percentages matter?
Yes, and this instrument cannot see it — a Net Promoter Score of 50 built from 10 responses and one built from 10,000 responses enter the two fields identically, even though the smaller sample can swing by dozens of points on the next wave purely from noise. Track the response count alongside the score itself, not just the score alone, before treating any change between two surveys as a real shift in loyalty.
Can a rising promoter share and a falling score happen together?
Yes, if the detractor share rises faster than the promoter share does. Because NPS subtracts one percentage from the other, a survey wave can show more promoters in absolute terms while still posting a lower overall score, if a support outage or a price increase pushed a larger group of former passives into the detractor bucket at the same time. Check both input percentages, not just the final score, whenever a reading moves.
How is the 0-10 recommendation question usually worded?
Most surveys use some version of Reichheld's original wording: how likely are you to recommend this company, product, or service to a friend or colleague, scored 0 (not at all likely) to 10 (extremely likely). Changing the wording, the scale, or the timing relative to a purchase or support ticket can shift results enough that scores from differently worded surveys should not sit on the same chart.
References
- Harvard Business Review — Reichheld, 'The One Number You Need to Grow' (2003)
- U.S. Small Business Administration — Marketing and sales guide
Read this first: This instrument shows arithmetic, not advice. Real offers add fees, taxes and terms that vary by lender and place — verify the figures against your actual paperwork before deciding anything.