Last updated: 7 October 2026
Quick Answer: An NPS score, or Net Promoter Score, rates customer loyalty from one 0 to 10 question asking how likely someone is to recommend you. Subtract the percentage of detractors (0 to 6) from the percentage of promoters (9 or 10). Scores run from -100 to +100, and Fred Reichheld of Bain introduced the metric in 2003.
An NPS score is the share of customers who would actively recommend you minus the share who would steer others away, taken from a single 0 to 10 question. Two businesses can post the same +20 for opposite reasons: one has 20% promoters and no detractors, the other 60% promoters and 40% detractors, a split base where four customers in ten are unhappy.
That compression is the metric's strength and its blind spot. One net figure is easy to track by team and quarter, but it hides the mix and every reason, and at the sample sizes most programs field, a 10-point move between waves can be noise.
How Is NPS Calculated?
To calculate NPS, sort every 0 to 10 answer into three bands, then subtract the share of detractors from the share of promoters. Promoters answer 9 or 10, passives 7 or 8, and detractors 0 to 6. Passives count in the total but in neither group, so a large passive share pulls the score toward zero.
The net promoter score formula is NPS = % promoters - % detractors. Here it is for 200 responses, rows running from the top of the scale down.
| Band | Answers | Responses | Share of total |
|---|---|---|---|
| Promoters | 9 or 10 | 90 | 45% |
| Passives | 7 or 8 | 70 | 35% |
| Detractors | 0 to 6 | 40 | 20% |
| NPS | 200 | 45 - 20 = +25 |
The NPS calculation takes four steps:
- Drop blank and "don't know" answers, then count the valid responses.
- Count the responses that fall in each band.
- Divide each band by the valid total to get its share.
- Subtract the detractor share from the promoter share.
What Does the NPS Score Range From -100 to +100 Tell You?
The NPS score range runs from -100, when every respondent is a detractor, to +100, when every respondent is a promoter. Any positive score means promoters outnumber detractors. Zero is not neutral, though. It can come from a base that is entirely passive or from 40% promoters facing 40% detractors.
The cutoffs are deliberately harsh: a 6 sounds like a pass, yet it counts against you. Bain, which co-owns the method, says detractors account for more than 80% of negative word of mouth and promoters for more than 80% of referrals, while passives repurchase and refer at rates up to 50% lower than promoters.
What Is the NPS Question, and How Should an NPS Survey Ask It?
The NPS question asks how likely it is that a person would recommend a company, product or service to a friend or colleague. They answer on an 11-point scale labeled 0, not at all likely, to 10, extremely likely. A standard NPS survey then adds one open follow-up asking why, which is where the usable information sits.
Small wording choices decide whether the score holds still from wave to wave:
- Fix the stem: Asking about the last order instead of the brand creates a new metric, not a new reading.
- Ask it early: Satisfaction items placed first can prime the answer, much as leading question patterns do.
- Pick relational or transactional: Relational NPS asks about the whole relationship on a schedule; transactional NPS asks right after a purchase or support contact.
Where Did the Net Promoter Score Come From, and Does It Predict Growth?
Fred Reichheld of Bain & Company introduced the Net Promoter Score in a December 2003 Harvard Business Review article, "The One Number You Need to Grow." He argued that one recommend question tracked growth better than long satisfaction surveys. Independent tests back a narrower version of that claim.
A 2007 Journal of Marketing study by Timothy Keiningham and colleagues re-ran the analysis on 21 firms and 15,500-plus interviews from the Norwegian Customer Satisfaction Barometer. It could not replicate the "clear superiority" of NPS over other measures, including the American Customer Satisfaction Index.
A 2022 study in the Journal of the Academy of Marketing Science tracked seven US sportswear brands for five years. Only the brand-health version of NPS, asked of all potential customers rather than current ones, predicted future sales growth. Who you ask decides what the score can forecast.
How Many Responses Does an NPS Score Need to Be Reliable?
An NPS built on 200 responses carries a margin of error of roughly 11 points either way at 95% confidence when promoters and detractors are both common. Each answer scores +100, 0 or -100, which makes NPS noisier than a single percentage. Halving the margin takes four times the sample.
For the 45% promoter, 20% detractor mix above, the standard error is the square root of (P + D - (P - D) squared) divided by n, with P and D as proportions, assuming a random sample.
| Responses per wave | Margin of error on one score | Change needed between two waves |
|---|---|---|
| 100 | ±15.0 points | 21.2 points |
| 200 | ±10.6 points | 15.0 points |
| 400 | ±7.5 points | 10.6 points |
| 1,000 | ±4.8 points | 6.7 points |
Segment cuts are worse: a store with 30 responses can sit 27 points either side of its true score.
Which Customers Does an Email NPS Survey Miss?
An email or web-link NPS survey hears mostly from customers who open email, read the survey language comfortably and like answering surveys. Disengaged detractors and customers who prefer another language are the likeliest gaps, and the score can rise just because those people stop answering.
Language is a bigger gap than most programs assume. In 2019, 67.8 million US residents spoke a language other than English at home, almost one in five, according to the Census Bureau.
Mode matters too. In a Pew Research Center experiment with 3,003 panelists randomly assigned to phone or web, answers differed by 5.5 points on average, and phone respondents reported higher satisfaction with family and social life by 18 and 14 points. Keep the mode fixed across waves.
Changing the channel changes who answers. WhatsApp-native interviews put the recommend question and its follow-up inside a chat the customer already uses, with no link and no app, and accept voice notes. AI phone interviews reach customers who take calls but ignore email. Alchemic publishes 60+ languages including Spanish, Hindi and Arabic, with managed fieldwork or bring your own sample across 14 markets including the USA and the UK.
Where Does NPS Fall Short?
NPS falls short as a diagnostic: it shows how the balance of promoters and detractors moved, not why. Use it as a headline tracker inside a voice of customer program, paired with open follow-ups and experience questions.
Four limits recur:
- It discards information. A 0 and a 6 count the same, and so do a 9 and a 10.
- It summarizes experience loosely. In patient surveys at six Dutch hospitals, with 6,018 inpatients and 10,902 outpatients, NPS tracked reported experiences less closely than a global rating.
- Scale use varies by culture. Response styles differ across countries, as Anne-Wil Harzing's 2006 review documents, so a multi-market NPS mixes loyalty with scale habits.
- Incentives inflate it. When the score drives frontline bonuses, the pressure to ask customers for a 10 is built in.
The fix is pairing, not replacement. Deeper Brand Tracks adds 40 to 80 AI-moderated interviews to each tracker wave, so an NPS dip in one segment arrives with its reasons the same week.
Once it has the client's brief, Alchemic designs and tailors the discussion guide to handle these risks before fielding, rather than leaving that work to the buyer. For a quick pulse on a digital product with thousands of users, an in-app NPS prompt from a survey tool is cheaper and usually enough.