What is Net Promoter Score (NPS)?

Net Promoter Score (NPS) is a customer loyalty metric that measures the likelihood a customer will recommend a company, product, or service to a friend or colleague, expressed as a single number on a scale from -100 to +100. It is collected with a one-question survey — "how likely are you to recommend [company] to a friend or colleague?" — answered on a 0-to-10 scale, and calculated by subtracting the percentage of detractors from the percentage of promoters.

NPS has been the most-cited customer experience metric for the past two decades. It is short enough to embed almost anywhere, simple enough for a board to understand, and contested enough in academic literature that any serious user has to know the limitations as well as the formula.

This page covers what NPS measures, where the methodology came from, how to calculate it, who counts as a promoter, what a "good" score looks like by industry, how NPS compares to CSAT and CES, where the metric is strong, where it falls short, and a short FAQ.

NPS in one sentence

NPS is the percentage of customers who are likely to recommend a brand, minus the percentage who are unlikely to recommend it.

What NPS actually measures

NPS is a relationship metric. It is meant to capture how a customer feels about a brand or product overall — not how they felt about one specific interaction. The question is broad on purpose: would you tell a friend to buy this? Anyone who answers that question is, in effect, putting their social capital on the line, which is what gives the response its weight.

That scoping is the whole point. NPS is designed to give a leadership team a single number that travels — across boards, investor calls, annual reports, and quarterly reviews — and represents the durability of the customer base. If the number is going up, customers are increasingly willing to vouch for the brand. If the number is going down, it can be an early signal that the customer relationship is weakening before churn shows up in the revenue line.

Customer loyalty as a concept is broad and fuzzy. NPS as a metric is narrow and headline-friendly. The narrowness is a feature for executive reporting and a bug for diagnostic work — which is why most mature CX programs run NPS alongside other metrics rather than alone.

Where did NPS come from?

NPS was introduced by Fred Reichheld, a director emeritus at Bain & Company, in a December 2003 Harvard Business Review article titled "The One Number You Need to Grow". The methodology was developed by Reichheld in collaboration with Bain & Company and the analytics firm Satmetrix.

The origin story Bain tells is specific. Reichheld and a Bain team ran a research project with data supplied by Satmetrix, testing a wide range of survey questions to find the single best predictor of customer behaviors that drive growth — repurchases and referrals. They tested questions like "how satisfied are you?", "does this company deserve your loyalty?", and "do you intend to return?" against actual purchase data. The "how likely are you to recommend" question outperformed the others in 11 of the 14 industries they studied, and was close to the top in the rest.

The score itself is a registered trademark. Net Promoter®, NPS®, and the related emoticons are jointly owned by Bain & Company, NICE Systems, and Fred Reichheld. The "℠" service mark applies to the score itself, and Bain has since extended the methodology into what it calls the Net Promoter System® — the broader operating model for collecting, acting on, and rewarding NPS data inside a company.

NPS scaled fast. As of 2020, Fortune reported that two-thirds of the Fortune 1000 had adopted some version of NPS, and the methodology is now embedded in the standard customer-experience playbooks of nearly every large B2C and B2B operator.

How NPS is measured

An NPS survey has three moving parts: the question, the scale, and the formula.

The question

NPS is a single-question survey. The wording is fixed:

"How likely are you to recommend [company / product / service] to a friend or colleague?"

The simplicity is the point. The standard practice — recommended by Bain and built into nearly every NPS platform — is to follow the rating with one open-ended "why?" question, which is where most of the diagnostic value actually lives.

The scale

NPS uses an 11-point scale, 0 to 10, where 0 means "not at all likely" and 10 means "extremely likely." Respondents are then bucketed into three groups, as Bain documents on its Net Promoter System site:

• Promoters (9–10). Loyal enthusiasts. Bain's own data shows promoters account for more than 80% of referrals in most businesses, and have substantially higher repurchase rates than passives.

• Passives (7–8). Satisfied but unenthusiastic. Their repurchase and referral rates are as much as 50% lower than promoters', and they are the most likely group to defect to a competitor.

• Detractors (0–6). Unhappy customers. Detractors account for more than 80% of negative word-of-mouth and have high rates of churn.

The 0–10 range, rather than a 1–5 or 1–10 scale, is deliberate. Bain argues the 11-point scale produces measurable behavioral differences between adjacent score bands, and avoids the default-to-the-middle problem that shorter scales create.

The formula

The accepted NPS formula is:

NPS = % promoters − % detractors

Passives are not counted directly in the calculation, but they are not ignored — they pull down the percentage of promoters in the denominator.

The result is reported as an integer, not a percentage. The range runs from -100 (every respondent is a detractor) to +100 (every respondent is a promoter). A score of zero means the brand has as many people willing to actively recommend it as it has actively warning friends off.

Worked example: 100 customers respond. 60 give a 9 or 10 (promoters), 20 give a 7 or 8 (passives), 20 give a 0 to 6 (detractors). NPS = 60 − 20 = +40.

Relational vs. transactional NPS

Two deployment models are common.

Relational NPS surveys are sent at a regular cadence — quarterly, semi-annually, annually — and ask about the overall relationship. This is the version NPS was designed for.

Transactional NPS surveys are sent right after a specific interaction — a support contact, a purchase, an onboarding step. Many programs run these, but Qualtrics explicitly recommends against the practice, arguing that NPS measures a relationship and applying it to single transactions misuses the instrument. The more conventional metrics for transactional measurement are CSAT and CES.

NPS vs. CSAT vs. CES

NPS is one of three single-question metrics that show up in almost every customer experience program. They measure different things, ask different questions, and are useful in different moments.

The shortest way to keep them straight: NPS is about the relationship, CSAT is about the moment, CES is about the friction. Most mature CX programs run all three because each one answers a question the others cannot. For a deeper read on how these metrics work together inside a contact center, see Gladly's guide to contact center metrics for modern CX teams.

A note on overlap: NPS and CSAT are sometimes used interchangeably because both fire post-purchase, but they are not the same instrument. CSAT asks how the customer felt about the experience; NPS asks whether the customer would refer a friend. A customer can be satisfied with a specific interaction and still be unwilling to recommend the brand — which is why running both produces a more complete picture than running either one alone.

What is a good NPS score?

There is no universal "good" NPS, because expectations vary by industry, region, and survey methodology. Bain's own scoring framework, as referenced by Qualtrics, gives the most-cited general anchors:

• Above 0 — good. More promoters than detractors.

• Above 20 — favorable. The brand has clear advocates.

• Above 50 — excellent. The brand is in the top tier of its category.

• Above 80 — world-class. Rare and difficult to sustain.

Industry-specific numbers from Retently's 2026 NPS benchmark study give a more granular read. Recent industry averages include:

• Consulting and digital marketing agencies: 83 (top of the published range)

• Financial services: in the high 60s, leading the B2B pack

• B2B software and SaaS: mid-40s

• Construction: rebounded to 42

• Cloud and hosting: 39

• Internet software and services: 26

• Communication and media: 26 (bottom of the published range)

Three rough bands hold up across most data sets:

• Below 0 — net detractors. More customers actively warning others off than recommending. Needs immediate diagnosis.

• 0 to 30 — neutral to mildly positive. Most customers are passive. Stable but not durable.

• 30 to 70 — healthy range. This is where most well-run B2C and B2B programs operate.

• Above 70 — exceptional. Often achievable only in concentrated categories (luxury retail, niche SaaS, premium consulting) and very rarely sustained across an entire customer base.

The more useful question than "is my NPS good?" is "is my NPS moving in the right direction?" A score that climbs five points a quarter usually means more than a one-time score that is high.

What are the strengths of NPS?

NPS has dominated CX dashboards for twenty years because the strengths are practical.

It travels well. A single integer between -100 and +100 fits on a slide. Executives, investors, boards, and frontline managers can all read it the same way. Few other CX metrics travel the same distance with the same fidelity.

It is cheap to deploy. One question, any channel. NPS can be sent over email, SMS, in-app, post-purchase, post-renewal, or embedded in a help center. The implementation cost is low and the data starts arriving the day the survey goes live.

It correlates with growth. Bain's own analysis argues that sustained value creators have NPS roughly twice as high as the average company, and that NPS leaders grow at more than twice the rate of competitors. The strength of the correlation is contested in academic work, but the practical pattern is recognizable inside most companies running the metric.

It is benchmarked. Bain's NPS Prism, the Qualtrics XM Institute, Retently, and CustomerGauge all publish industry data. That makes peer comparison possible without expensive primary research.

It pairs well. NPS, CSAT, and CES are complementary, not competitive. A program that runs all three sees both the moment-level signal (CSAT, CES) and the relationship-level signal (NPS) without either one fighting the other.

What are the limitations of NPS?

NPS is one of the most-criticized metrics in customer experience research, and any serious user has to know the limitations as well as the formula.

It is a single-question survey. As the Wikipedia entry on NPS notes, the validity and reliability of any company's NPS depend on collecting a large number of responses. Academic critics have argued that the "likelihood to recommend" question is not a meaningfully better predictor of business growth than other customer-loyalty questions, including straightforward satisfaction and repurchase questions.

It tells you what, not why. NPS produces a number. It does not produce a reason. The follow-up "why?" question is what does the diagnostic work, and many programs collect the score without doing anything with the open-text response — which leaves the metric as a thermometer with no thermostat.

It is biased toward extremes. The cleanest, most actionable signals on an NPS distribution are at the ends — the 9s, 10s, and 0s. Passives, who are the largest group in most distributions, are functionally invisible in the calculation, which is the group that defects fastest.

Reichheld has criticized its overuse. In a 2025 interview on WBUR's On Point, Reichheld himself described the over-deployment of NPS surveys as a "tragedy of the commons" — customers are now bombarded with so many requests for ratings that response rates have declined and the data quality has degraded. He has also publicly recommended de-linking NPS from employee compensation, arguing that compensation linkage makes employees care more about getting a high rating than about pleasing the customer.

"Likelihood to recommend" is not the same as "would recommend." NPS is a stated-intention metric, not a behavioral one. Customers who say they are highly likely to recommend a company often do not. Pairing NPS with actual referral data — measured by campaign attribution or self-reported "how did you hear about us" surveys — closes that gap and gives the relationship-level signal a behavioral anchor.

The practical takeaway: NPS is a strong default signal for the overall customer relationship, but it is one signal, not the whole picture. It tells the boardroom that something is up. It does not tell the operating team what to do about it.