Net Promoter Score ranks customers on a single question — "How likely are you to recommend us to a friend or colleague?" — on a 0-to-10 scale, then nets the enthusiastic responses against the unhappy ones into one number between -100 and 100. It became one of the most widely adopted customer metrics in business, pitched from the start as something more direct than a satisfaction survey: a single score that would tell a management team, at a glance, whether the company was actually going to grow.
Unlike some of the folklore we've traced elsewhere in this series, NPS has a real, identifiable creator and a real, dated origin story — no fabricated inventor here. What's disputed isn't who made it. It's the flagship claim that made it famous: that this one number is "the single most reliable indicator of a company's ability to grow." That claim has since been tested, adversarially, by researchers publishing in some of the most rigorous marketing journals in the field — and the replications and rebuttals that followed tell a far messier story than the one-number pitch.
Key takeaways
- NPS's origin is genuine and well documented: Fred Reichheld introduced it in a December 2003 Harvard Business Review article, building on a study of eight survey questions tested against the future purchasing and referral behavior of 4,000 consumers across 14 case-study companies.
- The scale's cutoffs aren't an even split — Promoters score 9-10, Passives 7-8, Detractors 0-6. Reichheld's own account says the lines were drawn where referral and repurchase behavior actually changed in his data, not by dividing 0-10 into equal thirds.
- Two independent, peer-reviewed studies in top marketing journals — one of which won an award for "the most significant contribution to the advancement of the practice of marketing" — directly tested Reichheld's growth claim and found no support for it, including a case where a standard satisfaction index beat NPS at predicting growth in two of the three industries Reichheld himself used to showcase it.
- Byron Sharp, director of the Ehrenberg-Bass Institute for Marketing Science, called the original analysis "sloppy" and its resulting reputation "snake oil, fake science" — and a separate, independently-run academic study found the Customer Effort Score, often marketed as NPS's methodologically superior replacement, actually performed worse than both NPS and plain satisfaction at predicting real, measured customer retention.
- There's no single "good NPS score." An independent consumer panel puts the software industry's average NPS at 15.6; benchmark reports built from software vendors' own self-selected customers put it at 29 to 41 — a gap of more than 2.5x for the same label, depending entirely on who was sampled.
What is the Net Promoter Score, and how is it calculated?
Every respondent answers one question on an 11-point scale (0-10): how likely are they to recommend the company to a friend or colleague. Their answer sorts them into one of three groups, and the score nets two of those groups against each other.
NPS = %Promoters − %Detractors, reported as a plain integer from -100 to 100 — not a percentage, even though it's built out of two of them. A company where 60% of respondents are Promoters and 10% are Detractors scores +50, regardless of how the remaining 30% (Passives) answered.
- Promoters (score 9-10): loyal enthusiasts who keep buying and refer others.
- Passives (score 7-8): satisfied but unenthusiastic, and considered vulnerable to a competitor's offer.
- Detractors (score 0-6): unhappy respondents who can damage a brand through negative word of mouth.
Where did NPS actually come from, and what did its creator originally claim?
Fred Reichheld, a Bain & Company consultant, introduced the metric in "The One Number You Need to Grow," Harvard Business Review, December 2003 — a real, dated, easily verified origin, unlike some of the manufactured folklore we've run down elsewhere on this site. His account describes testing eight candidate survey questions ("How satisfied are you?", "Does this company deserve your loyalty?" and others) against the actual future purchasing and referral behavior of 4,000 consumers across 14 case-study companies. The "would you recommend" question came out the best or second-best predictor of that behavior in 11 of the 14 cases.
A second, larger analysis ran in parallel, conducted by the measurement firm Satmetrix: over 50 companies across roughly a dozen targeted industries, correlating each company's average NPS for 2001-2002 against its own revenue growth for 1999-2002. That's a design worth sitting with — it compared historical NPS to historical growth over an overlapping window, not a score measured at one point in time against growth that hadn't happened yet. Reichheld expanded the argument into the book The Ultimate Question (2006), and again into The Ultimate Question 2.0 (2011, with Rob Markey), which reframed NPS as part of a broader "Net Promoter System."
"Net Promoter", "NPS" and "NPS Prism" are registered trademarks, and "Net Promoter Score" and "Net Promoter System" are service marks — currently held jointly by Bain & Company, Fred Reichheld personally, and NICE Systems (which acquired Satmetrix, the metric's original co-owner, in 2017-18), per Bain's own trademarks-and-licensing page. It's a live, actively licensed mark, not an informal industry term anyone owns outright.
Why the odd scale — why does a 6 count as a "detractor" and not just "okay"?
The 9-10 / 7-8 / 0-6 split looks lopsided next to an even three-way division of a 0-10 scale, and Reichheld's own explanation is that it isn't arbitrary: he reports drawing the lines where referral and repurchase behavior actually changed in his data, not where the numbers happened to divide evenly. On his account, respondents rating 9 or 10 were the ones who referred and repurchased at rates high enough to visibly move a business; a rating as "middling" as 6 behaved, in practice, much closer to a 0 or a 1 than to a 9.
An independent replication effort run by the UX-research firm MeasuringU (Jeff Sauro and James Lewis) tested a version of this directly rather than taking it on faith. Across a sample of about 500 customers, they found Detractors produced roughly 90% of the negative comments in their data — not identical to Reichheld's own figure, but pointing the same direction. In a separate, larger longitudinal sample (n=6,026), the share of actual recommendations that came from Promoters ranged from 51% to 77% depending on the product category — a real, meaningful skew toward the top of the scale, even if it's short of a clean "80% of everything."
Does NPS actually predict revenue growth better than other metrics?
Bain's own current claim, published on its Net Promoter System site, is specific: "in most industries, Net Promoter Scores explained roughly 20% to 60% of the variation in organic growth rates among competitors," and "on average, an industry's Net Promoter leader outgrew its competitors by a factor greater than two times." That's the pitch that made NPS famous — a single customer-survey number that, on its own, explains a large share of who grows and who doesn't.
The most direct test of that pitch came from Timothy Keiningham, Bruce Cooil, Tor Wallin Andreassen and Lerzan Aksoy, "A Longitudinal Examination of Net Promoter and Firm Revenue Growth," Journal of Marketing, Vol. 71 (July 2007) — a paper that, somewhat ironically, went on to win the Marketing Science Institute's 2007 H. Paul Root Award for "the most significant contribution to the advancement of the practice of marketing." Using 21 Norwegian firms and more than 15,500 customer interviews from the Norwegian Customer Satisfaction Barometer, plus a digitized reconstruction of Reichheld's own published charts, the authors tested whether NPS actually behaved the way the pitch describes.
| Claim you'll see repeated everywhere | What the peer-reviewed research found |
|---|---|
| NPS is "the single most reliable indicator of a company's ability to grow" | Keiningham et al. (2007) tested this directly and concluded: "We find no support for the claim that Net Promoter is the 'single most reliable indicator of a company's ability to grow.'" |
| Customer satisfaction (like the ACSI) has no real link to growth — NPS is the superior alternative | Using Reichheld's own showcase industries, the same paper found the ACSI's R² actually beat NPS's R² in two of the three U.S. industries Reichheld used to demonstrate NPS's advantage. |
| An industry's NPS leader reliably outgrows competitors, because the underlying link is well established | A 2021 systematic review found only four academic studies have ever directly tested the NPS-to-growth link — and none of them confirmed Reichheld's original claim that NPS is a superior predictor of sales growth. |
Wintel personal computers
U.S. life insurance
Airlines
A second, independent study — Morgan & Rego, "The Value of Different Customer Satisfaction and Loyalty Metrics in Predicting Business Performance," Marketing Science 25(5), 2006 — reached the same bottom line from different data (ACSI, COMPUSTAT and CRSP figures, 1994-2000, across six metrics): "metrics based on recommendation intentions (net promoters) and behavior... have little or no predictive value... recent prescriptions to focus customer feedback systems and metrics solely on customers' recommendation intentions and behaviors are misguided." One honest complication worth flagging: Keiningham and colleagues later published a formal rebuttal disputing whether Morgan & Rego had actually measured NPS correctly (Marketing Science 27(3), 2008) — so the two "anti-NPS" papers don't fully agree on methodology, even though both independently failed to confirm Reichheld's claim.
Is NPS "junk science" — or is that overstated too?
The sharpest-worded critique came from Byron Sharp, director of the Ehrenberg-Bass Institute for Marketing Science, who described Reichheld's original analysis as "sloppy" and, in a widely quoted 2008 remark, called the resulting metric "snake oil, fake science," adding: "the lesson for market researchers and insight directors is just how easy it is to make compelling slogans from incorrect findings." A separate, more clinically worded paper — Kristensen & Eskildsen (2011) — tested whether the promoter/passive/detractor cut-points hold up statistically and concluded: "the best we can say about NPS is that it is a mistake!"
A more measured academic critique comes from David Grisaffe, "Questions About the Ultimate Question: Conceptual Considerations in Evaluating Reichheld's Net Promoter Score (NPS)," Journal of Consumer Satisfaction, Dissatisfaction and Complaining Behavior, Vol. 20 (2007). His objection isn't that recommend-intent is meaningless — it's that collapsing a genuinely multidimensional idea like customer loyalty into a single 0-10 item, then inferring forward-looking, causal advice ("track this one thing to grow") from what's really a retrospective correlation, asks one simple number to do more conceptual work than a single number can bear.
It's worth adding a counterweight that's easy to miss in the pile-on: a separate, well-cited study — East, Hammond & Wright, International Journal of Research in Marketing, 2007 — found positive word of mouth actually outnumbers negative word of mouth by roughly 3 to 1 across the 15 product/service categories they examined. That cuts against the intuition (baked into NPS's asymmetric scale and its framing of Detractors as the loud, dangerous minority) that dissatisfied customers dominate the conversation. The strongest version of the critique isn't "recommend-intent tells you nothing" — it's that "the one number you need" oversells what any single-item survey metric can responsibly claim.
Is trust in NPS actually eroding among practitioners?
There's a real, dated, quantified signal here, not just an online mood. In a May 2021 press release, the research firm Gartner predicted that more than 75% of organizations offering CX (customer experience) technology would abandon Net Promoter Score as a key metric by 2025 — a forecast built on Gartner's own March 2021 survey of 42 customer-service leaders, 58% of whom said NPS had been mandated on their team from the top down rather than chosen for its usefulness.
Later reporting suggests the prediction landed roughly on target rather than as hype: NPS reportedly fell from the 2nd to the 8th most-used customer-experience metric between 2023 and 2024, and a 2025 TELUS Digital/Statista survey found only 23% of U.S. enterprise CX leaders still using it. At the product level, Zendesk quietly discontinued its own built-in NPS survey add-on — sales stopped in January 2023, and the feature was removed that April. None of this proves NPS is useless; it's a genuine shift in how much weight practitioners are willing to put on it as the metric.
23%
the share of U.S. enterprise CX leaders who told TELUS Digital/Statista's 2025 survey they still use NPS — down after Gartner's 2021 forecast that more than three-quarters of CX organizations would abandon it as a key metric by 2025.
Source: Gartner (May 2021 press release, n=42 CX leaders); TELUS Digital/Statista (2025)
So what actually counts as a "good" NPS score?
This is where the benchmark data itself becomes a cautionary tale. "What's a good NPS for my industry" gets answered very differently depending on who ran the survey and, specifically, whose customers were sampled.
The Qualtrics XM Institute's approach is the most methodologically independent of the sources we checked: it surveys ordinary consumers about named brands directly, explicitly excluding its own vendor's customer base from the calculation. Its Q3 2023 U.S. Consumer Benchmark Study — 10,000 consumers rating 351 companies across 22 industries — put Software at 15.6, roughly in the middle of the pack (the 22 published industry scores run from -6.1 for Consumer Payment up to 30.1 for Grocery, and average almost exactly 18 across the full list). Compare that to benchmark reports published by NPS-survey vendors themselves, built entirely from companies that already chose to run an NPS program on that vendor's platform: Survicate's 2025 report (599 companies, 2,187 surveys, 5.4 million responses) puts B2B software's median NPS at 29; Retently's most recent benchmarks (drawn from its own customer base, 10,000+ surveys) put B2B software and SaaS at around 41.
The likely explanation for the gap: vendor benchmark reports only ever see companies sophisticated (or invested) enough to already be running an NPS program on that vendor's software — a self-selected group that skews toward companies already paying attention to customer experience. An independent panel that asks ordinary consumers about brands they didn't choose to be surveyed about has no such filter. Retently itself has flagged a related problem worth citing: across its own data, survey response rates fell from a median of 32% in 2020 to 18% in 2025, and the customers most likely to bother answering skew toward the very happy and the very angry — inflating both Promoter and Detractor counts at the expense of Passives. Treat any single "good NPS is X" claim as meaningless without knowing exactly whose customers were asked.
Is there a better single metric — like the Customer Effort Score?
The Customer Effort Score (CES) was introduced as a direct challenger: Matthew Dixon, Karen Freeman and Nicholas Toman, "Stop Trying to Delight Your Customers," Harvard Business Review, July-August 2010, based on research at CEB (later acquired by Gartner in 2017) covering more than 75,000 customer interactions. The original question asked customers to rate, on a 5-point scale, how much effort they personally had to put in to get their issue resolved; a 7-point "the company made it easy for me to handle my issue" agreement statement is the more common version sold today.
The specific number that gets repeated constantly — that CES is "1.8 times more predictive of loyalty than CSAT and 2 times more predictive than NPS" — traces only to marketing blogs quoting each other. We could not find it in any accessible CEB or Gartner report, methodology note, or primary source; treat it the same way we'd treat any other widely copied statistic with no traceable original. What we did find is a real, independent, peer-reviewed test that ran all three metrics head-to-head against something CES was specifically supposed to beat NPS and CSAT at: actual measured customer behavior, not self-reported intent.
| Feedback metric | Correlation with actual 2-year customer retention |
|---|---|
| Top-2-box customer satisfaction | r = .184 (best of the five metrics tested) |
| Official Net Promoter Score | r = .170 |
| Average satisfaction score | r = .151 |
| "NPS value" (a satisfaction-weighted variant) | r = .159 |
| Customer Effort Score | r = −.073 — the only metric pointing the wrong direction |
de Haan, Verhoef & Wiesel, "The predictive ability of different customer feedback metrics for retention," International Journal of Research in Marketing 32(2), 2015, tracked 93 firms across 18 industries and matched customer feedback scores against their actual retention two years later — not intention, the real outcome. Top-2-box satisfaction and official NPS came out statistically tied for the strongest predictor (Akaike weights of roughly 49.6% and 49.1% respectively, of being the single best metric), while CES was the best predictor in zero of the 18 industries tested and was the only metric of the five whose correlation ran in the wrong direction. It's a genuinely awkward finding for a metric marketed explicitly as NPS's methodologically superior replacement — and a reminder that "a metric with a research paper behind it" and "a metric that has been independently, adversarially retested" aren't the same claim. For what it's worth, CSAT itself has no single standard version across vendors; the one satisfaction metric with real academic pedigree behind it is the American Customer Satisfaction Index, founded in 1994 by Claes Fornell at the University of Michigan as a full structural model, not a bare survey score.
How should you actually use a recommend-intent score, if at all?
- 1Treat any single-item score — NPS, CES or a plain 1-5 satisfaction rating — as one input among several, not a replacement for looking at what customers actually do next (repurchase, churn, referral behavior you can track directly).
- 2If you run an NPS-style survey through a form tool or a dedicated survey platform, feed the resulting score onto the contact record as a custom field rather than leaving it stranded in a separate export — that's what makes it usable for segmentation and automation later.
- 3Branch on the score range rather than treating every response the same: a Detractor and a Promoter warrant completely different next steps, not the same generic "thanks for your feedback" reply.
- 4Route the two ends of the scale to different outcomes: a low score to a task for a human to follow up on before the relationship is lost, a high score to a referral or review-request flow while the customer is still enthusiastic.
- 5Re-ask on a schedule and watch the trend, not the snapshot — a single quarter's number, from a self-selected group of people who bothered to respond, is exactly the kind of small, noisy sample the peer-reviewed replications above kept warning about.
In oToK, a field-changes trigger can start an automation the moment a score lands in a contact's custom field — however you collected it — and a Multi-split decision node can route Detractors, Passives and Promoters down three separate branches in the same flow: a task for a teammate on the low end, a WhatsApp or email nudge toward a review or referral on the high end, tagging the contact automatically either way. An API step can also pull the score in directly from an external survey tool the moment a response comes in, so nothing sits in a spreadsheet waiting to be acted on.
Frequently asked questions
- Is NPS the same thing as customer satisfaction (CSAT)?
- No. NPS asks specifically about likelihood to recommend, while CSAT asks about satisfaction with a specific interaction or with the company overall — a related but conceptually different question. The peer-reviewed research comparing them found satisfaction-based metrics were, if anything, slightly stronger predictors of actual customer retention than NPS, not weaker.
- Who owns the NPS trademark, and can I use the term freely?
- "Net Promoter", "NPS" and "NPS Prism" are registered trademarks, and "Net Promoter Score"/"Net Promoter System" are service marks, currently held jointly by Bain & Company, Fred Reichheld, and NICE Systems (which acquired Satmetrix, the original co-owner, in 2017-18). You can describe your own recommend-intent survey using the term, but it's a live, actively licensed mark rather than free-to-claim generic terminology.
- Does a high NPS guarantee faster revenue growth?
- No. Bain's own materials describe a correlation, not a guarantee, and even that correlation has been directly disputed: peer-reviewed replications found NPS underperformed a standard satisfaction index in two of the three industries Reichheld used to showcase it, and a 2021 systematic review found no peer-reviewed study has ever confirmed Reichheld's original claim that NPS is the single best predictor of growth.
- What's a realistic "good" NPS score for a SaaS or software company?
- It depends entirely on who's asking. An independent consumer panel (Qualtrics XM Institute, 2023) put the software industry's average at 15.6; benchmark reports built from vendors' own self-selected customers put B2B software at 29 to 41. Compare your score against a source that discloses its sample and methodology, not against a single number quoted without either.
