What Is Net Promoter Score (NPS)? A Complete Guide to Measuring Customer Loyalty

Net Promoter Score is a single value to provide a snapshot of the true customer experience that you have. It has been in the boardroom decks since 2003, included in virtually every customer experience platform available and remains one of the more misunderstood metrics in business. Some teams see it as a Holy Writ. Others dismiss it as a vanity number and state that its significance is limited. The truth lies between: NPS is really useful, but only if you know what it is, why it is wrong, and what you’re supposed to do with this number when you have it.

This guide explains all that what NPS is, how to determine it properly without making common pitfalls, what a good score really means when you take your industry into account, where it can mislead you, and how you can have a program around it that actually changes customer behavior rather than just producing a chart to be placed on a slide.

What Net Promoter Score Actually Measures

NPS is a question: How likely is it that you would recommend this company or product to a friend or colleague on a scale of 0 to 10? Nothing else. There was no survey with 20 questions, nor the need for a research department to interpret it.

From that one answer, all the respondents are divided into three groups. Your customers who are actually loyal to you and will refer people when you ask them are your promoters, who answer 9 or 10. At 7 or 8, they are content enough to remain but not so enthusiastic that they will drum up interest, nor so excited that they’ll leave while another competitor makes a better offer. Anybody under 6 will be unhappy enough that they will actively dissuade others from working with you.

Your Score = Promoter percentage – Detractor percentage. A question like “likelihood to recommend” is not particularly good because of the arithmetic; it’s because it’s a more powerful predictor of future actions such as repurchases, referrals, and “sticking around” versus churning than is the simple question of satisfaction with the last interaction. Satisfaction is backward-looking. NPS is attempting to forecast the next action of someone.
Part of the reason that the metric spread so quickly is that of that predictive angle. It’s fast enough to keep the response rate at a decent level, easy enough for a founder or a front-line manager to read without having to train them, and has been around for long enough that there are actual benchmark data to compare against.

Where NPS Came From (and Who Owns the Name)

Net Promoter Score was conceived by Bain & Company partner Fred Reichheld in a Dec. 2003 Harvard Business Review article, “The One Number You Need to Grow.” The research team at Reichheld had been testing dozens of survey questions against their actual repurchase, referral and long-term retention rates among customers across a broad spectrum of industries to look for the one that predicted growth best. The recommendation question was successful in every aspect, which led to the creation of a whole framework by Reichheld around it, expanded in his book The Ultimate Question, which was published in 2006.

If you spend a lot of time reading official material on this, then you will need to know that Net Promoter, Net Promoter Score and NPS are registered trademarks of Bain & Company, Fred Reichheld and Satmetrix. The term is now used generically, as with “Google it,” but it doesn’t have the same standing as “customer satisfaction”.

Since the original article, NPS has gained an employee-facing sibling (eNPS, as mentioned below), an extended tail of CX software designed to gather it, and more recently, AI capabilities to accelerate the processing of open text feedback. The core question hasn’t changed at all.

How to Calculate Your NPS

It’s that straightforward: NPS = promoters percentage – detractors percentage. This is where people get the subtraction wrong for the first time when they do it by hand: passives count as part of your total respondents, but not towards the subtraction itself.

Here is a clean sample. Suppose that you conduct a survey of 200 customers. 120 of them answer 9 or 10 that’s 60% promoters. 50 answer 7 or 8 25% are passives. The other 30 answer 6 or less than 15% detractors. Your NPS is 60 minus 15, or 45.

Be aware that the passive group is not removed from the picture when it is not in the formula. Many teams overlook this 25% completely; a large number of this group are the most cost-effective to move to the promoter column, as they are already happy and may only require one more reason to get excited.

For how the number indicates that you’re in more critics than fans territory: anything below zero is a true alarm signal, not a rounding error. Scores 0-30 are acceptable but can be made better. At 30, you’re pretty good for most businesses, and at 50, you’re beating the majority of your competition. Anything over 70 is rare and will likely be among companies that have a really excessive amount of attention to product or service quality, like the Apples and Costcos of the world. All this is relative and not definitive, as what’s good is significant depending on what you’re selling.

What Counts as a Good Score Depends on Your Industry

It could be a good score for a Telecom or Utility firm and a bad score for a Software company because the expectations when customers walk into these firms are totally different. On average, those industries involved in day-to-day exposure to friction such as utilities or traditional telecom have lower scores, whereas industries such as specialty retail or well-run SaaS products tend to score meaningfully higher. B2B companies as a whole tend to have a more concentrated score range, in part because there are fewer touchpoints in a B2B relationship, and in part because a displeased B2B customer is more likely to raise a problem than to simply walk away and leave a low rating unspoken.

But none of this means don’t worry about your number being less than a SaaS benchmark you read in a blog post. It means the thing you need to compare is your direct competitors and against yourself a year ago. A growth in a category with a low average score is more significant than a flat score in a category with a high average score.

But that is only one input in a broader competitive gap analysis, which is a more systematic approach to understanding which areas of our competitors’ business are gaining traction and which are not, as a bigger whole that requires a greater number of metrics beyond just loyalty.

Why NPS Correlates With Revenue and Retention

Much of the reason that NPS became more than just a curiosity in the world of customer service is the research that Bain did. One of the firm’s findings, with its co-author Earl Sasser of Harvard, is that increasing customer retention by 5% can boost profitability by 25% to 95% in various industries due to the lower cost of retaining a customer compared to acquiring a new one. Companies with higher and better NPS scores also appear to be among the fastest-growing in their industry, to a large extent because promoters purchase more frequently, remain more engaged and recommend others at close to no acquisition cost.

It’s worth being honest about what that does and doesn’t prove. Just a higher score is not enough to generate more revenues on its own, and there’s plenty of NPS marketing material that furthers the line. The truth, though, is that businesses with truly good customer relationships have faster growth a slightly different, more defensible proposition, and one that’s the real reason to make a program.

Where NPS Falls Short

There’s no metric so plain that after two decades of use, it doesn’t get some legitimate criticism from real people, and NPS has received some of this.

The scoring bands lose real information. It’s not like you’re dealing with the same level of grievance from a customer that gives you a 0 and one that gives you a 6. It is the same with promoters, a 9 and a 10 are both considered the same.

It is the easiest of all the companies to game. Making the score known to the customer in person immediately following a positive interaction, or letting them know that the employee’s bonus is tied to their response, will actually inflate the number, but not move real loyalty. Retail and field-service environments are plagued with the issue all the time, when a salesperson gives a tablet to someone as he or she walks out the door.

Does not cross-culturally well. It’s not that countries don’t report satisfaction in the range of 0 to 10, some countries are more likely to assign a 9 or 10 to someone who is happy, and some countries are less likely to assign a 9 or 10 to someone who is happy. Raw scores will give you a false impression of which market is the happier one if you compare them without taking into account that tendency.

And only reaches people who are still here. A relationship survey predominantly includes customers who remained to obtain it. Lots of the most useful people to learn from were never in the sample anyway.
These all don’t render NPS useless. This makes it a warning and not a finding. The programs that deliver real value from it are the ones that show how the scores correlate with behavioral trends in customer usage of the programs, the volume of customer support requests, and when they signed up for the next quarterly program.

NPS, CSAT and CES are not competitive metrics

The three of them are frequently grouped together, and it is assumed that you should choose one and ignore the others. They address different questions in practice and the best CX programs execute all three.

NPS is practiced at the relationship level. Asked quarterly or annually, it’s the one to use when you want a single trend line to report to the board or when comparing to a competitor. CSAT works at the moment level. It asks you immediately after a certain interaction, such as a purchase, a support ticket or a demo, and is focused on what just happened, not the customer’s overall experience.

What CES measures is effort, not sentiment. There is one that is most helpful of these three, and it’s the checkout flow, which is too long, or the onboarding flow that confuses people, or the support flow that requires five clicks when it should only require one.

Together, they tell a fuller story than any one could tell alone. NPS tells you loyalty is slipping somewhere. CSAT provides you with information about the interaction that it is associated with. Most often, CES will provide you with an answer: that too much effort was expended throughout the process. Teams that monitor just one of the three are more often than not half-informed.

Relationship NPS vs. Transactional NPS

The second difference that is as important as the metric you use is often overlooked: it’s a relationship survey vs. a transactional survey.

Relationship NPS is a measure of the health of the relationship. It is released on a regular basis quarterly, semi-annual, sometimes annually and is the one that is included in board reporting since it helps to track the trend of loyalty over time, no matter what the customer has been doing lately.

Transactional NPS occurs after the customer has just completed a transaction: bought a product or service, made a support call, reached a milestone in the onboarding process, or renewed. It records the feel of that particular moment when it is fresh, which is much more useful to see the exact point where something is going wrong in a process.

Most experienced programmes use both, targeting different audiences within the programme. As a strategic health check, Relationship NPS goes up to leadership. Transactional NPS belongs to the teams that are responsible for each touchpoint, since it provides them with information they can use this week, not this quarter.

When and How to Actually Send the Survey

The question is not as important as people think if it comes out at the wrong time or isn’t in the right form. There are a few things that are always a part of an NPS program that result in useful data, and there are a few things that are always a part of an NPS program that result in noise.

Limit the survey to a few minutes. Almost all the time, this three-question approach core question and one follow-up open-ended question, what’s the main reason for your score? will generate a higher response rate than a long questionnaire, which results in a quick drop-off in response interest. That follow-up, honestly, is often more useful than the number itself. It’s where you find out what’s getting you and your customers more loyal, or more frustrated.

Measure it to the business, not the clock. SaaS businesses tend to receive more meaningful data after waiting 30 days or so until they’ve actually had a chance to experience the product after signing up, and then restating the same question every 3 months thereafter. It’s better shortly after delivery than when the product is actually in someone’s hands or when they check out. B2B-relationships are more resilient to timing based on milestones, onboarding, after renewal, and after a quarterly business review than a fixed date on the calendar. Hospitality and travel businesses get their most accurate read immediately after the stay or the flight, while the experience is still fresh.

Match the channel with the time. Email is more effective for the slower, considered, relationship-level survey, while in-app or on-page surveys tend to get a higher response rate because they are capturing someone at the moment.

And don’t over-ask. Repeating surveys to the same customers over the course of a few weeks yields a drop in response rates and less elaborate answers, which is typically referred to as survey fatigue. Rotating the questions asked if you have a large customer base in the spacing relationship survey and switching to a quarterly survey continues to provide meaningful data.

Closing the Loop on Feedback

The one habit that can have the greatest impact in any NPS program is closing the loop, and that’s the habit many companies are too embarrassed to practice after the initial excitement of the program.

The individual is the one who reads back what the person has written and responds: giving a thank you to promoters, a follow-up to detractors within a day or two instead of a week, and one clarifying question to passives to get to the bottom of what is holding them back from being enthusiastic. At the systemic level, it involves repeating and repeating the cycle of product and leadership with the same themes that arose from the feedback, and also stating clearly to customers when a change was made as a result of the feedback.

That second bit, when you tell customers what you changed as a result of their feedback, is essentially underpromise and overdeliver for feedback. You do not have to follow all of the suggestions. You’re performing more than the customer intended and not saying a whitethumb about it, and that is the sort of follow-through that gets the customer to trust you once more.

Earning a Higher Score without Gaming It

Achieving better NPS is typically a matter of straightforward, routine behaviors, not a single event.

Begin with the open text feedback, not the number. Learn what it is detractors are actually saying, identify 2 or 3 complaints that appear repeatedly and repair those issues before moving on with any new projects. Increase the speed and the authenticity of support many negative comments on ticket numbers are really about how they feel. Then provide an actual easy-to-understand way for your promoters to send business your way. There are lots of businesses with enthusiastic customers but no easy way to send them through.

This final one relates more than just to customer service. Genuine advocates are a hotter lead than pretty much any cold outbound list can generate, and are precisely the kind of lead that comes through carefully planned, systematic sales prospecting techniques.

When faced with limited time, a personal follow-up with a disgruntled customer is likely to shift the score more than a reward for a satisfied customer. Rewarding promoters is a nice thing to do, not a strategy. The actual power is in the hands of those who aren’t getting attention: the detractors and passives.

Employee NPS (eNPS): The Other Half of the Picture

The question Employee Net Promoter Score has is the same one asked within the company: On a scale of 0 to 10, how likely are you to recommend this company to others as a place to work? The same three groups will apply: engaged employees who would actively recommend the job (9-10), employees who are satisfied but not that excited about the position if they had a better one to go to (7-8) and disengaged employees who are a flight risk, who are also spreading negative sentiment that works its way out from the organization (0-6).

This isn’t an accident and is part of the same discussion as customer NPS. Happy staff members provide better service and it’s reflected in the scorecard of their customers. In short, a company that fails to account for the feeling of employees is not really capturing where the true culprit of a low NPS may lie. The issue may stem from a support team that is burnt out, understaffed, and/or not engaged long before a customer ever encounters an issue with their experience.

Where AI Actually Helps (and Where It Doesn’t)

The majority of AI’s contributions to NPS programs in 2026 have to do with speeding up tasks that previously took an analyst days to complete manually. Thousands of open-text answers are organized into themes, pricing, onboarding, a particular feature, and support quality, which was previously a week’s worth of work. When someone leaves a negative comment behind, but it is not urgent, the flag will help to demark it as such and prevent the most urgent feedback from getting lost in the spreadsheet until the monthly review. When NPS data is linked to product utilization and support interactions, there are some platforms that can alert to accounts that appear to be on the path to scoring poorly before the survey even gets sent.

All that doesn’t take the place of judgment, and it’s okay to be straight about that, rather than over-selling it. The sorting, tagging and flagging is the area where AI actually excels at scale that no human team could have achieved in the same time. What to do about any given detractor and what to do about it to follow up in a way that isn’t a form letter is still a human task. The programs that are actually benefiting from the AI-powered analysis are doing so to quickly identify the signal, before giving the real response to a human.

The Bottom Line

The reason NPS belongs in a CX program is that it’s not a perfect measure, but it’s so easy that any whole company can get behind it, and so specific that if you ask one more question, you know where to look up next. Don’t use it alone, use it with CSAT/CES and actual usage data. Be honest about the places where it can lead you astray. Use it as a way to have an actual conversation with that customer, whether it’s a positive or negative response.

If it’s your first time using this, build it up slowly: One clean question, one honest follow-up, and one actual person who will make contact with all of the low scores by the end of the week. Once that basic loop is running, everything else, like segmentation, dashboards, and AI-assisted theme extraction, can be added on.