Customer & product
Net Promoter Score
A customer loyalty metric built on one question, how likely are you to recommend us, scored 0 to 10. Subtract the percentage of detractors (0 to 6) from the percentage of promoters (9 and 10) for a single figure that is easy to track, easy to compare, and easy to abuse.
Also known as NPS, Net Promoter System. First set out by Fred Reichheld, with Bain & Company and Satmetrix in 2003; the primary source is cited in full below.
- Format
- Scoring model
- Level
- Corporate · Business unit · Product
- Best for
- Understand customers
- Decision stage
- Diagnose · Review
- Difficulty
- Introductory
- Time to apply
- A day to design and launch a sound survey; the closed-loop system is an ongoing operating rhythm.
Plate · The model
The components
Ask the likelihood-to-recommend question
The standard single question on a 0 to 10 scale, plus an open why. Its virtue is comparability and low friction; its limit is that one question carries all the weight, so wording, timing and sampling discipline matter more than they would in a longer instrument.
Signals of strength
Standard wording and scale, unchanged between waves · Sample covers the silent and the unhappy, not just engaged customers · Survey timing is consistent and not cherry-picked after good experiences
Segment promoters, passives and detractors
The three bands: promoters at 9 and 10, passives at 7 and 8, detractors at 0 to 6. Passives count in the denominator but not the score, which is easy to forget and means a wave of 7s moves the score without a single detractor.
Signals of strength
Bands applied exactly, with no local reinvention of the cut-offs · Detractor comments read individually, not just averaged · Distribution reported, since the same score can hide very different mixes
Compute promoters minus detractors
The arithmetic that produces the headline number, from minus 100 to plus 100. Differencing two percentages discards information and inflates sampling error, so small movements between waves are usually noise.
Signals of strength
Sample size and confidence interval reported with the score · Wave-to-wave changes tested before being celebrated · Relationship and transactional scores never mixed in one trend line
Close the loop with follow-up
The system around the score: rapid contact with detractors, root-cause work on recurring themes, and easy referral paths for promoters. Reichheld's own position is that the score without the follow-up system achieves nothing.
Signals of strength
Detractors contacted within days by someone empowered to fix things · Recurring themes traced to operational changes with owners · Front-line teams see the verbatims, not just the number
When it earns its keep
- You want a simple, repeatable pulse on customer loyalty that non-specialists across the organisation can understand and rally around.
- You need a consistent measure to track over time or across branches, products or regions, where the trend and the spread matter more than the absolute number.
- You want a systematic trigger for follow-up conversations: the score identifies who to call, and the verbatim comments tell you what to fix.
- Referral and word of mouth genuinely drive growth in your market, which is where the recommendation question has the most claim to relevance.
And when it doesn't
- As the single measure of customer health. The evidence that NPS uniquely predicts growth has not survived independent replication; treat it as one indicator among several.
- When the score is wired to bonuses or league tables. Once pay depends on it, staff learn to beg for nines and tens and the number stops measuring anything.
- In business-to-business settings with a handful of accounts, where a sample of six responses makes the score statistically meaningless and a structured account review tells you more.
- When you cannot resource the follow-up. A score collected without closing the loop is survey theatre, and customers notice being asked and ignored.
How to run it
Before starting, gather the inputs the analysis depends on:
- A defined survey approach: relationship NPS on a regular cycle, transactional NPS after key moments such as installation or a service visit, and the discipline not to muddle the two.
- A clean, unbiased sampling frame covering detractors as well as the customers most likely to respond warmly.
- The standard 0 to 10 question wording, plus an open follow-up question asking the reason for the score, which is where the actionable content lives.
- An owner and a process for closing the loop with respondents, especially detractors, within days rather than quarters.
- 1
Ask the likelihood-to-recommend question
Put the standard question, how likely is it that you would recommend us to a friend or colleague, on a 0 to 10 scale, to a representative sample. Pair it with one open question asking why. Keep the survey short; response rate is part of the measurement.
- 2
Segment the responses
Classify 9 and 10 as promoters, 7 and 8 as passives, and 0 to 6 as detractors. The bands are deliberate: Reichheld's argument was that only enthusiasm predicts advocacy, and a polite 7 is not loyalty.
- 3
Compute the score
Subtract the percentage of detractors from the percentage of promoters to give a score from minus 100 to plus 100. Report the sample size and response rate alongside it; a score of plus 40 from thirty responses is a hint, not a fact.
- 4
Read the reasons
Code the verbatim answers to the why question by theme. The score tells you the temperature; the comments tell you what is burning. Themes that recur among detractors are the operational priority list.
- 5
Close the loop
Contact detractors quickly to recover the relationship and learn the specifics, thank promoters and make referral easy, and feed the themes into operational fixes. This follow-up discipline is what Bain calls the Net Promoter System, and it is where the value sits.
- 6
Track the trend, protect the measure
Watch movement over time and variation across units rather than chasing an absolute benchmark. Keep the metric away from individual bonus formulas and audit for score-begging, or the trend you are watching will be the gaming, not the loyalty.
Reading the result
A headline score with its distribution and sample size, a themed digest of the reasons behind it, and a closed-loop action list: which detractors were recovered, which root causes were fixed, and how the trend moved by segment.
- Read the trend and the spread, never the absolute number in isolation. Scores vary by industry, country and survey method, so cross-company comparisons are mostly noise.
- Treat the verbatims as the primary output and the score as the index to them. A rising score with worsening detractor themes is a sampling problem, not an improvement.
- Check the mechanics before believing any movement: response rate, sample mix and any change in when or how the survey was asked.
A worked example
A domestic boiler installation and servicing firm puts NPS to work
A regional UK boiler installation and servicing firm with 30 engineers runs on referrals and repeat annual service plans, so recommendation is commercially real for it. It introduces transactional NPS after installations and service visits, and a relationship survey to its service-plan base each spring, after a winter in which complaints rose and two engineers left.
- Ask the likelihood-to-recommend question
- Text message with the standard 0 to 10 question and one open why, sent two days after each job to every customer, not just installations that went smoothly. Response rate 34 per cent. The firm resists the office manager's suggestion to skip customers who had complained, which would have quietly gutted the sample.
- Segment promoters, passives and detractors
- Spring wave, 412 responses: 58 per cent promoters, 27 per cent passives, 15 per cent detractors. Installations skew strongly promoter; annual service visits carry nearly all the detractors. The passives cluster around one theme, courteous engineers but vague arrival windows and slow quotes for follow-up work.
- Compute promoters minus detractors
- Headline score plus 43, with a confidence interval of roughly nine points either way. Installation NPS is plus 68; servicing NPS is plus 21. The firm decides the split, not the headline, is the finding, and declines to put plus 43 on the website next to competitors' unverifiable numbers.
- Close the loop with follow-up
- The operations lead phones every detractor within three days. The calls surface a pattern the themes had hinted at: missed service appointments cluster on two overbooked rounds, and detractors were rebooked weeks out. Scheduling is rebalanced, a same-week rebooking rule is set for missed visits, and promoters get a referral card with their service certificate. Engineers see the verbatims monthly; nobody's pay is tied to the score.
The read. Six months on, servicing NPS moves from plus 21 to plus 38 while installation NPS is flat, which is what the firm should expect given where it acted. The honest reading is that NPS did not tell the firm anything its complaint log could not have, but it forced a cadence: every unhappy customer got a call, every theme got an owner, and the recommendation question fitted a referral-driven business. The managing director's instinct to set engineers a target of plus 50 was the one decision the framework's own evidence base argues hardest against, and it was dropped.
Pitfalls
- Chasing the number instead of the reasons. The score is an index; the verbatims and the closed-loop fixes are the product.
- Tying the score to individual bonuses, which reliably produces score-begging, sample-gaming and pleading footers on invoices, and destroys the measure's information content.
- Comparing your score with published competitor numbers collected by different methods, at different touchpoints, in different countries.
- Reading noise as signal. Differencing percentages inflates sampling error, so a five-point wobble between waves on a few hundred responses usually means nothing.
- Surveying without acting. Customers who score you a 2 and hear nothing have now been let down twice.
- Ignoring cultural and channel response styles: some markets rarely award a 10 and phone surveys score differently from text, so mixed-method trends mislead.
What the critics say
The founding empirical claim failed replication. Using longitudinal data from the Norwegian Customer Satisfaction Barometer across 21 firms, Keiningham and colleagues could not reproduce the asserted link between Net Promoter and revenue growth, and found NPS no better than traditional satisfaction measures as a growth predictor, directly contradicting the 'one number you need' claim.
Keiningham, T. L., Cooil, B., Andreassen, T. W. and Aksoy, L. (2007) 'A Longitudinal Examination of Net Promoter and Firm Revenue Growth', Journal of Marketing, 71(3), pp. 39-51.
In practice the metric is widely gamed. A Wall Street Journal investigation documented NPS cited roughly 150 times in a year of S&P 500 earnings calls, tied to executive bonuses, collected by staff coaching customers to give tens, and reported without disclosing method, turning a customer-listening tool into a self-graded vanity statistic.
Safdar, K. and Pacheco, I. (2019) 'The Dubious Management Fad Sweeping Corporate America', The Wall Street Journal, May 2019.
Methodologically, collapsing an 11-point scale into three bands and differencing two percentages throws away information and inflates the variance of the estimate, so NPS needs larger samples than a mean score to detect the same change, while the 9-10, 7-8 and 0-6 cut-offs rest on one US cross-industry study and travel poorly across cultures with different response styles.
A second replication reinforced the first. Using Dutch data across business and consumer markets, van Doorn and colleagues found NPS neither superior nor inferior to conventional satisfaction and loyalty metrics: all performed equally well on current margins and revenue growth, and all performed equally poorly at predicting future sales growth, margins and cash flows.
van Doorn, J., Leeflang, P. S. H. and Tijs, M. (2013) 'Satisfaction as a predictor of future performance: a replication', International Journal of Research in Marketing, 30(3), pp. 314-318.
Grisaffe's conceptual review catalogued the measurement problems: a single item cannot carry the construct's weight, stated intention to recommend is not actual recommending behaviour, the promoter and detractor labels misdescribe many respondents, and the aggregate score conceals the segment-level differences a firm would need to act on. The 'ultimate question' framing, he argued, encourages firms to stop measuring precisely what they need to understand.
Grisaffe, D. B. (2007) 'Questions about the ultimate question: conceptual considerations in evaluating Reichheld's Net Promoter Score (NPS)', Journal of Consumer Satisfaction, Dissatisfaction and Complaining Behavior, 20, pp. 36-53.
A statisticians' review sixteen years into the metric's life concluded that NPS has not delivered its claimed benefits: the single-question instrument is poor market research design, the score is statistically inefficient relative to the information collected, and value-based approaches that ask what drives customer choice outperform it as a guide to action.
Fisher, N. I. and Kordupleski, R. E. (2019) 'Good and bad market research: a critical review of Net Promoter Score', Applied Stochastic Models in Business and Industry, 35(1), pp. 138-151.
Work it through
Enter how many responses fell at each score from 0 to 10. Promoters (9 to 10) minus detractors (0 to 6), as percentages of all responses, give the score. Report it with its sample size: a score from thirty responses is a hint, not a fact, and the verbatim reasons, not the number, are the real output.
| Score | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Responses |
Promoters (9–10)
0 (–)
Passives (7–8)
0 (–)
Detractors (0–6)
0 (–)
Net Promoter Score · 0 responses
–
Promoters minus detractors, as a share of all responses. Report it with the sample size: a score from thirty responses is a hint, not a fact, and the reasons behind the scores are the real output.
Sources and further reading
- Reichheld, F. F. (2003) 'The One Number You Need to Grow', Harvard Business Review, 81(12), December 2003. ↗
- Reichheld, F. F. (2006) The Ultimate Question: Driving Good Profits and True Growth. Boston: Harvard Business School Press.
- Keiningham, T. L., Cooil, B., Andreassen, T. W. and Aksoy, L. (2007) 'A Longitudinal Examination of Net Promoter and Firm Revenue Growth', Journal of Marketing, 71(3), pp. 39-51. ↗
- Safdar, K. and Pacheco, I. (2019) 'The Dubious Management Fad Sweeping Corporate America', The Wall Street Journal, May 2019.