What Is a Good NPS Score? First Ask How It Was Measured
A good NPS is above 0, with 30+ good and 50+ excellent, but the real answer depends on your industry and region. See the benchmarks and how to read them.
TL;DR
- The convention holds that above 0 is net positive, above 30 good, above 50 excellent and above 70 exceptional. Those bands are industry practice, not a Numr finding.
- They are also close to meaningless on their own, because the same company can produce two different scores without its service changing at all, purely from how, where and when the question was asked.
- The level of a score is a product of the method. Its movement is the part that carries information. A company at 25 and climbing is in better shape than one at 55 and sliding.
- So the useful question is not whether your number clears a band. It is whether the number is going up, measured the same way as last time.
- The version of NPS that predicts customer behaviour is the relationship-level score, asked away from any single transaction.
The general benchmarks
If you want the quick reference, these are the bands most practitioners use:
NPS range | What it signals |
|---|---|
Below 0 | More Detractors than Promoters. Something needs investigating |
0 to 30 | Net positive. Acceptable, with clear room to improve |
30 to 50 | Good. A healthy base of Promoters |
50 to 70 | Excellent. Strong word of mouth driving growth |
70+ | Exceptional. Rare, and usually a market leader |
Those bands are the start of the answer, not the answer. Two things decide what your number actually means: your industry, and, less discussed, your measurement method. The industry effect is real and widely written about. The measurement effect is larger than most teams realise, and almost nobody publishes numbers on it. We do, below.
Your method moves the score more than your service does
Across roughly 70 to 80 studies Numr has run spanning industries and touchpoints in India, Southeast Asia (Indonesia, Thailand, Singapore), Japan, Korea, and parts of Europe, one pattern repeats: scores collected by phone run 15 to 30 points higher than the same company measured digitally, depending on the industry and the study. The pattern holds across every market we have measured it in. What changes is the size of the gap, and it runs widest in Asia.
The mechanism is courtesy. A phone survey is a human talking to a human, and respondents, especially in India and Southeast Asia, do not want to appear impolite to the person on the line. They tend to appease, and the score carries that. The effect exists in Europe and the Americas too, a pattern Numr CEO Amitayu Basu saw first-hand over twenty years in CX, including at Satmetrix in the US, the company where NPS was co-created. It is simply less pronounced there.
A telephonic NPS in the 80s is not a world-class score. It is a telephonic one. The same customers asked digitally will say something less flattering, and neither figure is the true one. They are two different instruments pointed at the same customers.
We have watched this play out inside a single company. An insurer's telephonic programme carried an indicative NPS in the 80s and 90s, collected by agents calling a capped sample. Measured digitally across every touchpoint on Numr's platform, the score was 63. Nothing about the service changed in between; the instrument did. That gap runs wider than the 15 to 30 points we see across studies, which is what a capped, hand-picked calling sample does on top of courtesy.
The temptation at that point is to ask which number was correct. It is the wrong question and it wastes a quarter. Neither is a truer reading of the same quantity, because they are not measuring the same way. What the insurer could do, once the method stopped changing, was watch the digital number move.
Who answers changes the score too
A second effect stacks on top of courtesy, and it has nothing to do with politeness: who bothers to respond.
Digital and indirect surveys, where nobody is talking to anybody, are answered disproportionately by the extremes: the very happy and the very unhappy. The indifferent, usually the largest group a company has, do not reply at all, which is why an online distribution comes back bimodal with a thin middle. Phone surveys, by contrast, pull in far more Passives, people who would never volunteer a response but will answer a direct human, which fills the middle back in.
Those two effects pull in opposite directions, and courtesy wins. Politeness inflates a phone score; the Passives phone collects pull it back toward the middle. The net is still that phone reads higher, which is why the direction of the gap is stable even though two mechanisms are working against each other.
The practical consequence: an NPS of 40 measured digitally in Jakarta and an NPS of 40 measured by phone in London are not the same number, and neither comparison to a global benchmark chart means much.
Four more things that move the score
Where the question sits. Ask the NPS question first. Asked after a run of drill-down questions, the score absorbs whatever those questions just made the respondent think about.
When it is asked. Timing changes the answer, and it changes it by more than most programmes assume. Ask within hours of a car service and the glow of a clean handover produces a high score. Ask the same customer two days later, after a rattle has appeared, and it does not. Neither answer is false; they are different measurements.
What it is attached to. A single transaction is the wrong place to ask about recommendation. Nobody recommends a bank because one teller interaction went smoothly; that is hygiene, not advocacy. Most companies run transactional NPS anyway, and it is normal practice, but our advice is to measure effort at transactions and reserve the recommendation question for the relationship. Relationship-level NPS, asked away from any single interaction, is the version that predicts what customers actually do next.
How many respond. There is no single acceptable response rate, because the number is the product of two things: how invested the customer is in the category, and whether something has just happened that they have an opinion about. Numr has counted above 50 percent on a survey sent shortly after a premium vehicle purchase, above 20 percent on vehicle servicing, 14 to 15 percent immediately after a contact centre call, around 8 percent after a flight, and around 3 percent on a generic relationship survey with no triggering event. Each of those is a single programme rather than an average, observed across automotive, aviation, contact centre and relationship survey work, and they are what particular touchpoints produced rather than rates anyone should promise you.
One exception is worth stating, because it is the strongest evidence in the set: the above-50 and the around-3 figures are the same automotive brand, the same customers and the same channel in the same quarter, with only the moment differing. Note also that the top four are all immediately post-event and still span above 50 percent down to around 8 percent, so timing alone does not explain the spread between them. How much the customer cared in the first place does.
Judge a rate against its own touchpoint and its own category. Nine percent is poor for a contact centre and strong for a relationship survey, and a rate that looks thin can still be sound if the people who answered look like the people who did not.
Why a good NPS is still relative to your industry
With measurement held constant, industry differences remain large. Retently's 2026 NPS benchmarks put financial services at an average of 68 while internet software and services sits at 26, a 42-point gap driven by how much choice, friction, and emotion each category carries. A 40 would be unremarkable in one industry and a standout in the other.
This is why Numr runs proprietary industry NPS benchmark studies, measuring real scores across sectors and regions rather than borrowing a global average. Our refreshed India and Southeast Asia benchmark studies are in progress, and this page will link them when they publish. Until then, read our earlier rounds as history rather than a current reading. Our study of the NPS of Indian banks was fielded in 2019, and it still shows how far apart sectors and brands can sit, but the scores themselves have moved since.
Before you compare your score to any benchmark
Run this checklist. The more items you fail, the more directional your number is.
- Is it a relationship-level score, asked away from any single transaction?
- Was the NPS question asked first, before any drill-down questions?
- Was it collected without a human administering it?
- Was the response rate judged against its own touchpoint, rather than a global average?
- Was the sample uncapped and not hand-picked, covering all eligible customers?
Fail one, and read your score with a correction in mind. Fail three, and the level is not interpretable against anyone else's. It can still be interpretable against your own, provided you keep asking the same way.
Amitayu Basu, CEO and Co-founder, Numr"Ask two questions before you ask whether your NPS is good: how was it collected, and who answered. We have watched companies move up to 25 points without their service changing at all. The method was the score."
Chase the trend, not the trophy
This is the whole of it, and it is worth stating flatly. There is no clean, comparable absolute score waiting to be uncovered by better measurement. The level is a joint product of the category, the method and who chose to answer, and two scores are comparable only when all three match, which they almost never do.
So hold the method still and ask one question of your own programme: is this number going up. A company at 25 and climbing is in better shape than one at 55 and sliding. A method that is consistent and slightly biased beats one that is truer and keeps changing, because a consistent bias cancels out of a trend and an inconsistent one does not. And the payoff for moving it is real: Fred Reichheld's Bain research, published with Harvard Business Review in 2000, found that a 5 percent increase in customer retention can raise profits by 25 percent or more, so converting Detractors matters more than the headline number.
A benchmark also does not tell you what to do. Improving a good score into an excellent one comes from closing the loop on feedback: finding what is creating Detractors and fixing it before those customers leave. For the mechanics of the metric, see what Net Promoter Score is and how to calculate it.
Frequently asked questions
What is a good NPS score?
By industry convention, above 0 is net positive, above 30 good, above 50 excellent and above 70 exceptional. Those bands are practitioner shorthand rather than a research finding, and they assume a cleanly measured relationship-level score. Numr's position is that the level is the weaker signal: the same company can move 15 to 30 points on method alone. A good score is one that is going up, measured the same way as it was last time.
Why did our NPS drop when we moved from phone to digital surveys?
Two mechanisms. Human-administered surveys collect politeness along with opinion, which pushes a phone score up. A digital survey has to be chosen, so the people who answer are the ones with something to say while the indifferent do not reply at all, which hollows out the middle. They work on the score in opposite directions and courtesy is the larger of the two. Across our studies phone runs 15 to 30 points higher than digital measurement of the same company, depending on the industry and the study. Neither number is the honest one. They are different instruments, and what you lost was comparability with your own history, not accuracy. Hold the new method still and the trend becomes readable again within a cycle or two.
Is a good NPS the same in every country?
No, and the reason is method as much as culture. Courtesy bias is stronger in some markets than others, so the gap between a phone score and a digital score is itself regional. Cross-country comparisons need the same method and a regional benchmark, not a global chart.
What response rate makes an NPS credible?
There is no universal floor, and any vendor quoting one number across your whole programme is averaging unlike things. Response rate is the product of how invested the customer is in the category and whether something has just happened. Numr has counted above 50 percent on a survey sent shortly after a premium vehicle purchase and around 3 percent on a generic relationship survey with no triggering event. Those two are the same brand, the same customers and the same channel in the same quarter, with only the moment differing. Compare your rate to the same touchpoint, not to a global average, and check who answered before you worry about how many.
Should NPS be measured on transactions or on the relationship?
The relationship-level score is the one that predicts behaviour. Transactional NPS is common practice and not wrong to run, but a single interaction is hygiene, not a reason to recommend, and effort is usually the better transactional question.
Net Promoter Score, Net Promoter, and NPS are registered trademarks of Bain and Company, Satmetrix Systems, and Fred Reichheld.
