Net Promoter Score℠ (NPS®) turns one survey question into one number. Fred Reichheld of Bain & Company set it out in Harvard Business Review in December 2003. In a loyalty program it reads how members feel. What they do next sits in the program’s own data, and the two only make sense read together.
What is Net Promoter Score?
Net Promoter Score measures how willing customers are to recommend a company. Bain calls the survey item behind it “the ultimate question” (Bain). Reichheld argued that people who recommend a firm put their own name on the line, so they do it only when they are intensely loyal (HBR, 2003). His question read: “How likely is it that you would recommend our company to a friend or colleague?”
The question came out of testing. Reichheld and a Bain team checked many survey questions against what customers later did. In 11 of the 14 industry case studies they compiled, no other question was as strong at predicting behavior (Bain).
How do you calculate Net Promoter Score?
Ask the question on a scale of 0 to 10. Then sort each answer into one of three groups (Bain):
- Promoters, 9 or 10. Bain says they make up more than 80% of referrals in most businesses.
- Passives, 7 or 8. Bain says their repurchase and referral rates run as much as 50% lower than those of promoters.
- Detractors, 0 to 6. Bain says they make up more than 80% of bad word of mouth.
The score is the percent of promoters minus the percent of detractors. Passives count in the total but in neither group, so they pull the score toward zero. The score runs from -100, when every answer is a detractor, to +100, when every answer is a promoter.
What is the difference between relationship and transactional NPS?
Relationship NPS goes out with no trigger, often once or twice a year, to read how customers feel overall. Transactional NPS, which Bain calls experience NPS, follows a specific action, such as a purchase or a call to service. Bain adds a third type, competitive benchmark NPS. It says relationship and experience scores should not be compared directly (Bain).
| Type | When it is asked | What it shows | The loyalty program version |
|---|---|---|---|
| Relationship NPS | No trigger, often once or twice a year | How customers feel about the brand overall | A yearly member survey, read by tier and by tenure |
| Transactional (experience) NPS | Right after a specific action | How one interaction went, while it can still be put right | A short survey after a redemption, a tier change or a service failure |
| Competitive benchmark NPS | By a third party, double-blind | Where the brand stands against rivals | One sample with members of rival programs and with people in no program |
Bain also warns about who answers. Customers who dislike doing business with you can decide your survey is not worth their time. A third party asking hears from both ends of the scale (Bain).
What is a good Net Promoter Score?
A good score is one above your direct rivals, measured the same way. In 2003, Reichheld put the median score across more than 400 companies in 28 industries at just 16%. That rested on about 130,000 responses gathered by Satmetrix. The firms with the most eager referrals, such as eBay, Amazon and USAA, scored from 75% to more than 80% (HBR, 2003).
Bain ties the score to growth through rank against rivals. In most industries, NPS explained roughly 20% to 60% of the differences in organic growth rates among competitors. On average, the Net Promoter leader in an industry grew more than twice as fast as its rivals. Bain says the link is strongest where customers have a real choice and can switch with ease (Bain).
What are the limits of Net Promoter Score?
Outside tests have not backed its boldest claim. A 2007 study drew on 21 firms and more than 15,500 interviews. It looked at the same industries Reichheld cited. It could not reproduce the “clear superiority” he claimed for Net Promoter over other measures (Keiningham et al., 2007). A 2006 study was blunter. In US data from 1994 to 2000, scores built on intent to recommend had “little or no predictive value” for later business results (Morgan and Rego, 2006).
Intent also runs ahead of action. Researchers asked customers of a telecom firm and a financial services firm if they would refer the company, then tracked what they did. At the financial firm, 68% said they would and 33% did. At the telecom firm, 81% said they would and 30% did (HBR, 2007).
Reichheld has named the misuse himself. In 2021 he and two Bain colleagues wrote that some firms tied scores to frontline bonuses. Staff then cared more about the score than about learning to serve customers better. They added that survey scores are “inherently soft” (HBR, 2021). Our guide to measuring customer loyalty pairs NPS with the behavior it leaves out.
How do loyalty programs use Net Promoter Score?
First, split the score by program status. Bain says the score can be tracked for each customer segment (Bain). In a program, the first cuts are members against non-members and each tier against the next. Part of any member gap comes from who joins. A Dutch grocery study allowed for loyal shoppers selecting themselves into programs. Membership still lifted share of wallet, but by a seventh of what a naive model showed (Leenheer et al., 2007).
Second, time the transactional survey to program moments. Ask right after a redemption, when the program pays out. Ask after a tier change in either direction. Ask after a service failure, while the member can still be won back. Bain’s stated best practice is to ask any one customer no more than once every three months (Bain).
Third, join every answer to the member record. Then watch what each group does over the next 12 months: how often they redeem and spend, and whether they stay active. Bain notes that one feedback request can draw response rates in the low single digits (Bain). The program’s own data covers every member, including the ones who never answer.
Airlines already report the two side by side. American Airlines posted a 5-point rise in NPS for the second quarter of 2026, in the same release that put AAdvantage enrollments up more than 30%. It also cut NPS by flight outcome: for on-time flights, NPS rose for the 15th time in 17 months (American Airlines, Q2 2026).
For illustration, a hotel program surveys 2,000 members after a stay. Of those, 900 answer 9 or 10 and 400 answer 0 to 6. The other 700 are passives. Promoters are 45% and detractors 20%, so the score is 25. Now join each answer to the next 12 months. Promoters book 4 stays each in that year. Passives book 3, and detractors book 1.5. Moving 100 detractors up to passive is worth 150 stays a year. Moving 100 passives up to promoter is worth 100. The score of 25 alone would never show which group to fund first.
How GRAVTY handles Net Promoter Score
NPS answers come from a survey. GRAVTY holds the behavior to read them against. Each earn, burn, tier change and offer response is kept as a timestamped event on the member’s record (retention economics). Member 360 pulls each member’s behaviors and transactions into one view, household included.
Agentic AI Compass, the GRAVTY agentic AI suite, brings program KPIs and customer sentiment into one place (Agentic AI Compass). A loyalty team can ask in plain words why a metric changed. It can compare results across any segment. Uploaded survey results then sit beside the program’s own KPIs, so a falling score in one tier can be read against that tier’s behavior.
Bain asks for this credit on any page that uses its marks. Net Promoter®, NPS®, NPS Prism®, and the NPS-related emoticons are registered trademarks of Bain & Company, Inc., NICE Systems, Inc., and Fred Reichheld. Net Promoter Score℠ and Net Promoter System℠ are service marks of Bain & Company, Inc., NICE Systems, Inc., and Fred Reichheld.
Frequently asked questions
Is Net Promoter Score the same as customer satisfaction?
No. NPS asks how likely a customer is to recommend the company. A satisfaction survey asks how content they are. In a 2006 study, average satisfaction scores did the best job of predicting business results. Scores built on intent to recommend did little or nothing (Morgan and Rego, 2006). Track both, then test each one against what members spend.
Why do scores of 7 and 8 not count toward NPS?
They do count, but only in the total, so they pull the score toward zero. Bain calls these customers passives: satisfied for now, with repurchase and referral rates as much as 50% lower than promoters. Bain warns that a rival’s ad can pull them away (Bain).
How often should a loyalty program survey members for NPS?
Run a relationship survey once or twice a year and a short transactional survey after key moments. Bain gives both cadences and advises asking any one customer at most once every three months (Bain). In a program, the moments worth a survey are redemptions and tier changes. Service failures belong on the list too.
Should NPS be tied to staff bonuses?
No, by the account of the metric’s own creator. Reichheld and two Bain colleagues wrote in 2021 that tying scores to frontline bonuses made staff care more about the score than about serving customers. They proposed earned growth as an accounting-based counterpart. It counts customers who came through a referral or recommendation as earned (HBR, 2021).
Is NPS a percentage?
NPS is a number from -100 to +100, built from two percentages. Reichheld’s 2003 article wrote scores as percentages, such as a median of 16% (HBR, 2003). Bain asks users not to write “NPS Score,” since that reads as “Net Promoter Score Score” (Bain, trademarks).