NPS stands for Net Promoter Score. It’s one question: “How likely are you to recommend us to a friend or colleague?” People answer from 0 to 10. You turn those answers into a single score between -100 and +100.
That’s the whole idea. The rest of this guide covers how to calculate it, how much to trust it, and what to do with it once you have it.
What NPS is
NPS sorts everyone who answers into three groups:
| Answer | Group |
|---|---|
| 9 or 10 | Promoters |
| 7 or 8 | Passives |
| 0 to 6 | Detractors |
Your score is the percentage of promoters minus the percentage of detractors. Passives count in the total, but they don’t add or subtract anything.
If everyone answers 9 or 10, you score +100. If everyone answers 6 or below, you score -100. Most real scores land somewhere in between.
The idea comes from Fred Reichheld, who described it in a 2003 Harvard Business Review article. The wording and the groups are standard, so a survey that uses them is comparable to your own earlier surveys.
How to calculate it
Say you got 200 answers:
- 100 gave a 9 or 10. That’s 50% promoters.
- 60 gave a 7 or 8. That’s 30% passives.
- 40 gave a 0 to 6. That’s 20% detractors.
Your NPS is 50 minus 20, which is 30.
You write it as 30, not 30%. It’s a score, not a percentage.
Notice what the passives did: nothing. And notice that a 6 counts exactly the same as a 0. If a detractor who gave you a 2 changes their mind and gives you a 6, your score doesn’t move.
Don’t average the 0 to 10 answers. That gives you a different number that isn’t NPS.
To skip the arithmetic, use the free NPS calculator. It also shows a confidence range, which matters more than most people expect. That’s the next section.
How to read the number
Your NPS is an estimate. You asked some of your customers, not all of them, so a different batch of respondents would give you a slightly different score.
The calculator shows how much. It gives a 95% confidence range: the score plus or minus 1.96 standard errors. In plain terms, it’s the span the true score probably sits inside.
Using the example above, 200 responses and a score of 30:
- The range is about plus or minus 11 points. Your real score is probably somewhere between 19 and 41.
With only 50 responses (20 promoters, 20 passives, 10 detractors), your score is 20. The range is about plus or minus 21. That’s anywhere from about -1 to 41.
With 800 responses and the same shares, the range drops to about plus or minus 5. Four times the responses cuts the range roughly in half.
This changes how you should read quarter-to-quarter moves. Say your score goes from 30 to 33 and each round had 200 responses. That 3-point gain is well inside the noise. It doesn’t tell you anything got better.
A few rules that follow:
- Compare against your own history, asked the same way, to the same kind of customers.
- Check the range before celebrating or panicking. A move smaller than the range is not news.
- Don’t chase a “good” number. Whether a score is good depends on who you ask and how. We cover that in what is a good NPS score.
SurveyRock’s NPS question type gives you the promoter, passive and detractor breakdown. It doesn’t show a confidence range in survey results. For that, use the calculator.
Relationship NPS and transactional NPS
There are two common ways to use the question.
Relationship NPS asks your customers how they feel about you overall. You send it on a schedule to a broad group, say every customer who has been with you three months or more. It answers: “How do people feel about us right now?”
Transactional NPS asks right after a specific moment: a support chat, a delivery, an onboarding call. It answers: “How did that go?”
Both use the same question. They tell you different things, so don’t mix them in one number. A relationship score of 30 and a post-support score of 30 are not the same measurement.
When NPS is the wrong tool
NPS asks about loyalty, not about a single experience. If you want to know whether one support ticket went well, a satisfaction question fits better. It asks “How satisfied were you with this?” and gives you a CSAT score. If you want to know how hard something was to do, ask an effort question.
NPS is also a blunt tool when you have very few customers. With 25 responses, the range is so wide the score can’t guide a decision. Read the comments instead.
The question and the follow-up
The standard wording is: “How likely are you to recommend [company or product] to a friend or colleague?” The scale runs from 0 (not at all likely) to 10 (extremely likely).
Keep it as written. Changing the words or the scale makes your new score hard to compare with your old ones.
Then add one follow-up: “What’s the main reason for your score?”
This is where most of the value is. The score tells you something moved. The comments tell you what. Ten detractors who all mention slow onboarding is a plan. A score of 22 is a number.
Keep the follow-up open-ended and make it one question, not three. A longer survey loses people before they get to it. The survey length guide covers how much a few extra questions cost you in finished responses.
Running it
Who to ask. Ask people who have actually used the product. For a relationship survey, that usually means customers past a first few weeks. For a transactional one, ask whoever just had the experience.
How often. For relationship NPS, quarterly is a sensible start. Twice a year works too. The deciding factor is time to act: if you can’t fix anything between rounds, asking more often just produces more numbers that don’t change.
How many responses. There’s no magic number, but the examples above show the trade. Under about 50, the score is too loose to steer by, so lean on the comments. In the hundreds, you can start reading changes. The sample size calculator helps you plan how many you need.
What not to do. Don’t ask only after good moments, like right after a successful delivery. That inflates your score. And send reminders to everyone who hasn’t answered, not just your most engaged customers. If only the engaged ones get a nudge, the unhappy ones drop out and your score looks better than it is.
Building it. You can start from the NPS survey template rather than a blank page. If you also want a satisfaction check next to it, there’s a customer satisfaction template.
Closing the loop
A score on its own doesn’t change anything. What matters is what you do on the Tuesday a detractor comes in.
Here’s a workable path.
1. Read the reasons, not just the number. If you get a lot of open-text answers, group them. In SurveyRock, text analysis can group answers into themes. AI theme detection is available on Growth and above. With a hundred comments, “slow onboarding”, “can’t export” and “price” tend to sort themselves quickly.
2. Pick the one or two themes worth fixing. Not all of them. The ones that come up most among detractors, or the ones you can fix fastest.
3. Give each one an owner and a date. This is the step that gets skipped. In SurveyRock you can turn a theme into an Action with an owner, a due date and a resolution note. Free accounts can have 3 open Actions at once.
4. Follow up with the person. If you want to call back one specific detractor, you can create a “Follow up” Action from their response. The Action doesn’t send anything itself, so a person on your team writes or calls. If you’d rather reply automatically to every low score, a workflow can email the respondent (Team plan and up). And if you ran the survey in anonymous mode, there’s no contact attached to the response, so there’s nobody to call back. Decide that before you send.
5. Check whether it moved. Once 30 new responses have come in, an Action shows how the number compares with where it started. It’s labelled “movement, not proof”. That’s the right way to read it. Scores drift for lots of reasons, and one fix rarely explains a change on its own.
If you have rating questions in the same survey, key driver analysis (Growth and above) can show which answers tend to move together with the score. Read that as “these go up and down together”, not “this causes that”. It tells you where to look first. It doesn’t prove what’s behind the score.
Common mistakes
Believing a 3-point move means something. This is the most common one. A score moves from 30 to 33 and the team treats it as progress. With a few hundred responses, that’s inside the noise. Check the range first. If the new score is inside the old range, say “no clear change” and keep going.
Comparing yourself to other companies. Different industries, different customers, different survey timing and different wording all change the number. Compare to your own earlier rounds.
Changing the wording between rounds. Even small changes to the question or the scale break the comparison with last time.
Only asking happy people. Surveys sent right after good moments, or only to engaged users, give you a nicer score and a worse picture.
Collecting and not acting. If nobody owns a detractor theme, the next round will show the same themes. People who took the time to explain a low score notice when nothing changes.
Treating all detractors as one group. A 6 and a 0 count the same in the score, but they’re different customers. A 6 is often fixable. A 0 often has a specific story behind it. Read both.
FAQ
Is NPS a good metric?
It’s a good starting point and a poor stopping point. One question is quick to ask and easy to track over time. But it’s one number, it’s an estimate, and it doesn’t tell you why. Pair it with the follow-up question and an owner for what you learn.
How often should we send it?
For relationship NPS, quarterly is a sensible start, or twice a year if you can’t act that often. For transactional NPS, ask right after the moment you care about. Ask only as often as you can act on the answers.
Can NPS be compared across companies?
Not reliably. Industry, customer type, timing and wording all change the score. Compare against your own history, asked the same way.
What is the difference between NPS and CSAT?
NPS asks how likely someone is to recommend you, so it measures overall loyalty. CSAT asks how satisfied someone was with a specific thing, so it measures one experience. Use NPS for the relationship and CSAT for a single interaction like a support ticket.
Where to go next
Work out your own score and its range with the free NPS calculator. Then start your survey from the NPS survey template. For the plain definition of the term, see the glossary entry.
Net Promoter®, NPS®, NPS Prism®, Net Promoter System®, and the NPS-related emoticons are registered trademarks of Bain & Company, Inc., NICE Systems, Inc., and Fred Reichheld. Net Promoter Score℠ is a service mark of Bain & Company, Inc., NICE Systems, Inc., and Fred Reichheld.