Net Promoter Score: What It Actually Measures and What To Do With the Number

NPS is one of the most widely used metrics in business and one of the most widely misunderstood. Here is what the number really tells you — and what it does not.

In 2003, a consultant named Fred Reichheld published an article in Harvard Business Review arguing that most customer satisfaction surveys were too long, too infrequent, and too disconnected from actual business outcomes. His alternative was a single question: On a scale of zero to ten, how likely are you to recommend us to a friend or colleague? He called the resulting metric the Net Promoter Score, and over the following two decades it became arguably the most discussed number in customer experience.

You will find NPS tracked at companies ranging from four-person startups to Apple. You will also find it roundly criticised by researchers who question its predictive power, and quietly gamed by operations teams who time surveys to catch customers at their happiest. The truth about NPS sits somewhere between the evangelism and the scepticism. It is a useful tool, with real limitations, that most organisations either ignore entirely or treat as more meaningful than it is.

This article tries to get past both of those failure modes.

How the score is calculated

The mechanics are simple. You ask customers to rate their likelihood to recommend on a zero-to-ten scale. Based on their answer, you sort them into three groups.

Promoters — those who score nine or ten — are, in theory, your enthusiastic advocates: the customers who actually recommend you, who expand their own usage, and who are least likely to churn. Detractors — zero through six — are dissatisfied customers who may actively warn others away from you. Passives, scoring seven or eight, are satisfied but not enthusiastic. They are vulnerable to a competitor's pitch.

Passives being excluded from the calculation is one of the more counterintuitive design choices in NPS. The reasoning is that passive customers do not meaningfully drive growth in either direction. Whether that is empirically true is debatable, but it is worth understanding because it means a shift from passive to promoter does not move your score at all — only detractors converting to promoters (or vice versa) actually matters.

What a good score looks like

This is where teams often get confused, because NPS varies enormously by industry. A score that would be cause for celebration in telecommunications — an industry with notoriously dissatisfied customers — would be cause for concern in consumer software, where expectations are higher and switching costs are low. Comparing your score to an industry average matters much more than comparing it to a general benchmark.

More important than any benchmark, though, is your own trend over time. A score of 28 that has risen from 14 over six months tells you something meaningful and positive. A score of 45 that has fallen from 62 tells you something is wrong, even if 45 looks respectable in isolation. The direction matters more than the absolute number, particularly for small teams that cannot yet afford a statistically robust sample.

The question that matters more than the score

The single biggest mistake organisations make with NPS is treating it as a destination rather than a starting point. They send the survey, collect the number, put it on a dashboard, and move on. The score sits there, slightly improving or slightly declining each quarter, without generating any change in how the product or service actually works.

The reason NPS was designed with a follow-up question is that the number alone tells you almost nothing actionable. It tells you whether your customers are, on aggregate, happy or unhappy. It does not tell you why, where specifically, or what to do about it. The follow-up question — typically something like "What is the most important reason for your score?" — is where the actual insight lives.

When you read through open-text responses at volume, patterns emerge that no numeric score would surface. You might find that detractors cluster around a single pain point — say, the onboarding process, or a missing integration — while promoters consistently mention a specific feature or the quality of your support team. That is information you can act on. The score told you there was a problem. The responses told you where to look.

When to send the survey — and when not to

Survey timing has an outsized effect on results, and teams that do not think carefully about this end up with data that flatters rather than informs. Send a survey immediately after a customer successfully completes something they wanted to do — a successful onboarding, a resolved support ticket, a completed purchase — and you will get scores that reflect that positive moment, not their overall relationship with you. Send it at an arbitrary interval after sign-up, or during a period when a known bug is affecting users, and you get something different again.

There is no universally correct cadence, but a few principles hold up. For B2B SaaS products, a quarterly survey sent to active users tends to work well — often enough to catch meaningful trends, infrequent enough that customers do not develop survey fatigue. For ecommerce and consumer products, a transactional trigger after purchase or delivery makes more sense. For service businesses, sending within a week of a completed engagement, while the experience is still fresh, is typically the most useful moment.

The segments that matter most

Aggregate NPS hides as much as it reveals. A single organisation-wide score averaging together your power users, your barely-active users, your enterprise accounts, and your free-plan users is almost certainly telling you something meaningless about all of them. The more valuable practice is to segment your NPS data and look at each group separately.

Power users — those who are in your product every day and using its core features — will almost always have higher scores than occasional users. That makes sense. What you want to know is why the occasional users are occasional, and whether those who scored low can tell you what would change that. Segmenting by plan type, by company size, by industry, or by how long a customer has been with you will typically surface very different populations with very different concerns.

The most actionable segment of all is often the passives: customers who scored seven or eight. They are not unhappy — they are just not enthusiastic. Something is holding them back from full advocacy. A direct conversation with a handful of passive customers asking "what would it take to move you from an eight to a ten?" is one of the highest-return customer research activities available to a small team, and it costs nothing but time.

Closing the loop: what to do after you collect responses

The phrase "closing the loop" in the NPS world means following up directly with customers who responded — particularly detractors — to acknowledge what they said and, where possible, to do something about it. This step is where most organisations fall down. The survey goes out, the data gets analysed, the score goes on the slide, and the individual customers who took the time to tell you something are never heard from again.

This is both a missed opportunity and, for detractors, a small act of disrespect. A customer who scores you a three and explains why in some detail has given you a gift. If no one ever acknowledges it, you have signalled that you were not actually interested in the answer — just the number.

For detractors, the goal of follow-up is not to convert them into promoters on the spot — it is to understand their experience more deeply and, where possible, to resolve whatever caused the dissatisfaction. Many detractors who receive a genuine, personal response from someone at the company will update their perception significantly, even if the underlying product issue has not yet been fixed. The act of being heard matters.

For promoters, a simple acknowledgement — a thank-you email, or a question asking if they would be willing to share their experience publicly — can turn passive advocacy into active referrals. Promoters who score you a nine or ten are already inclined to recommend you. A small nudge to actually do so, at the moment when their enthusiasm is measured, costs almost nothing and occasionally converts into a case study, a review, or an introduction to another potential customer.

The honest limitations of NPS

Any balanced treatment of NPS has to acknowledge the genuine criticisms, because some of them are serious. The most significant is the cultural variation problem. Research has consistently shown that willingness to give extreme scores — nine or ten versus seven or eight — varies significantly by country and cultural context. Customers in the United States tend to score higher across the board than customers in Germany or Japan, independent of their actual satisfaction. If you are running a global product in forty markets, a single NPS figure will be partially measuring cultural response styles, not just customer experience.

There is also the question of whether NPS actually predicts what it claims to predict. Reichheld's original research linked high NPS scores to revenue growth. Subsequent studies have found the relationship to be much weaker and more industry-dependent than the initial claims suggested. NPS does correlate with customer satisfaction in most studies, but whether it is a better predictor of business outcomes than other satisfaction metrics is genuinely contested.

None of this means NPS is useless — it means it is a blunt instrument being used in contexts that sometimes call for sharper tools. For a team that currently measures customer satisfaction with nothing at all, starting with NPS is a significant improvement. For a team that has been running NPS for years and wants to understand customer behaviour more precisely, layering in additional metrics like Customer Effort Score, churn prediction models, or product usage data will give a much richer picture.

Building an NPS practice from scratch

If you are starting from zero, the most important thing is to start simple and stay consistent. Choose one customer segment to survey first — your most active users, or your most recent cohort of new customers. Write a survey with the standard NPS question and a single open-text follow-up. Use any survey tool that can send email (Typeform, Google Forms, and a dozen others will work fine for a small team). Set a calendar reminder to run it again in the same way in three months.

Do not invest in dedicated NPS software until you have run at least two or three rounds manually and have a clear picture of how the data will actually be used. The risk is spending time configuring a platform before you have established any habits around reading and acting on the results. A spreadsheet and a shared folder of open-text responses, reviewed as a team every quarter, will teach you more in the first six months than any dashboard.

The questions to ask yourself after each round are always the same: What is the score, and which direction is it trending? What are detractors saying, and is there a pattern? What are promoters saying, and are we doing more of that? Did we follow up with respondents last time, and what did we learn from those conversations? If you can answer all four questions clearly every quarter, you have a functioning NPS practice — regardless of what tool you are using to run it.