Introduction
You want one number that tells you whether customers would recommend you. That is what the Net Promoter Score, or NPS, was built to give you. It is the most common way to track customer satisfaction and loyalty over time.
This guide covers what NPS is, where it came from, and the exact math behind the score. You will learn how to field the question, what sample you need, and what counts as a good result. You will also learn where NPS and CSAT measure different things, so you do not confuse the two.
What Net Promoter Score (NPS) Is
NPS measures how likely your customers are to recommend you. Fred Reichheld introduced it in a December 2003 Harvard Business Review article, "The One Number You Need to Grow" (reprint R0312C), developed with Bain & Company. Bain now reports that NPS is used by about two-thirds of the Fortune 1000 (Bain & Company, "Net Promoter 3.0").
The method rests on a single question:
"How likely is it that you would recommend [Company] to a friend or colleague?"
Respondents answer on a 0 to 10 scale. Bain sorts those answers into three groups.
Promoters score 9 or 10. They are pleased enough to recommend your brand to others.
Passives score 7 or 8. They got what they paid for and nothing more, so they carry little loyalty value.
Detractors score 0 through 6. They are disappointed, and Bain says they can harm your growth and reputation.
Those definitions come straight from Bain, a co-owner of the NPS trademark (Bain & Company).
The NPS Formula
The score is a subtraction. You take the percentage of promoters and subtract the percentage of detractors.
NPS = % Promoters − % Detractors
Passives count toward your total responses, but they drop out of the calculation. Bain excludes them because neutral customers have a near-zero effect on referral behavior and revenue growth. The result runs from −100 to +100. Every respondent is a detractor at −100. Every respondent is a promoter at +100 (Bain & Company).
Here is a worked example. Say 200 people answer. 90 score 9 or 10, 70 score 7 or 8, and 40 score 0 through 6. Promoters are 45 percent. Detractors are 20 percent. Your NPS is 45 − 20, which is 25. The 70 passives shaped the percentages but never entered the subtraction.
How to Measure NPS Step by Step
Fielding NPS is straightforward once the question and sample are set. Follow these steps in order.
Step 1: Ask the Standard Question
Use the standard wording without editing it. Changing the recommend question breaks comparability with your own past results and with any benchmark. Keep the 0 to 10 scale, and label 0 as "not at all likely" and 10 as "extremely likely."
Add one open-ended follow-up: "What is the main reason for your score?" The number tells you where you stand. The comment tells you why.
Step 2: Decide Relational or Transactional Timing
Choose when to ask before you launch. NPS is usually a relational metric, fielded periodically rather than after a single interaction. Qualtrics notes it is useful after big company changes, marketing campaigns, or a PR crisis, and it is typically not tied to any one interaction (Qualtrics, "CSAT vs NPS").
Decide on a cadence and hold it steady. Quarterly is a common starting point. Consistency across waves matters as much as any single reading.
Step 3: Choose Your Sample
You need enough respondents to trust the percentages. Market research firms recommend a minimum of 300 to 400 respondents for a defined audience, and 500 to 1,000 when you want to break results down by region, age, or segment (Galloway Research; Attest). Attest advises 1,000 as an ideal, with 500 as a reliable floor.
Smaller samples still produce a number. They just widen the margin of error, so small movements between waves may be noise rather than change.
Step 4: Field to a Representative Group
Your score is only as good as who answers it. Send the study to a representative slice of your actual customer base, not just your happiest accounts or your loudest ones. Skewed sampling inflates or deflates the score without telling you.
Run quality checks on the responses. Attention questions, speeding detection, and one-response-per-device controls keep bad data out of your dataset before you calculate anything.
Step 5: Calculate and Record
Tally each group. Convert promoters and detractors to percentages of total responses. Subtract detractors from promoters. Record the score, the sample size, the date, and the exact question wording every time.
That record is what lets you compare wave to wave with confidence. Detecting a real change between two waves takes more sample than measuring a single level. As a matter of arithmetic, the standard error of a difference between two scores is the square root of 2, about 1.4 times, wider than the error on either score alone. Plan your sample for the comparison you want to make, not just the snapshot.
Key Considerations for a Trustworthy NPS
NPS answers one question well: would customers recommend you. It does not tell you why on its own, and it does not measure satisfaction with a specific transaction. That is where customer satisfaction score, or CSAT, comes in.
CSAT is transactional. NPS is relational. Qualtrics draws the line directly: one measures satisfaction with a specific interaction, the other measures overall loyalty and likelihood to recommend (Qualtrics, "CSAT vs NPS").
The two use different questions and scales:
CSAT asks, "How would you rate your overall satisfaction with the [service] you received?" on a 1 to 5 scale.
NPS asks the recommend question on a 0 to 10 scale.
You calculate CSAT differently too. Divide the number of satisfied customers, those who score 4 or 5, by total responses, then multiply by 100 for a percentage (Qualtrics).
Use CSAT right after a specific interaction, such as a support ticket or a delivery. Use NPS periodically to track the overall relationship. They are complementary. You cannot convert one into the other, so run both when you want the full picture of customer satisfaction.
NPS Benchmarks and Rules of Thumb
A number alone means little without context. Bain's scoring framework, as reported by Qualtrics, gives general thresholds (Qualtrics, "Net Promoter Score: The Ultimate Guide"):
Above 0 is good.
Above 20 is favorable.
Above 50 is excellent.
Above 80 is world-class.
Treat those as broad guides, not a universal pass mark. What counts as good depends heavily on your industry.
Industry averages vary widely. In Qualtrics XM Institute's 2024 U.S. consumer study (n=10,000, 354 companies, 22 industries), industry NPS averages ranged from 15.8 for car rental to 34.3 for grocers (Qualtrics XM Institute, 2024). A 30 might be middling in one sector and strong in another.
So the useful benchmark is your own history and your direct peer group. Compare this quarter to last quarter, and compare yourself to competitors in your category. A rising trend against your own baseline tells you more than any single threshold.
Common NPS Mistakes to Avoid
The method is simple, which makes it easy to run badly. Watch for these errors.
Editing the question. Rewording the recommend question makes your score incomparable to benchmarks and to your own past waves.
Fielding only to happy customers. A convenience sample of your top accounts produces a flattering number you cannot trust.
Reading noise as a trend. With a small sample, a two-point move between waves may sit inside the margin of error.
Ignoring the follow-up comment. The score tells you where you stand. The open-ended reason tells you what to fix.
Confusing NPS with CSAT. They measure different things on different scales and answer different questions.
Changing methodology mid-program. Shifting the wording, scale, or sample source creates artifacts that look like real change.
Fix these at setup. Lock the question, define a representative sample, hold your method steady, and read the score against your own trend.
What to Do With Your NPS Results
A score is a starting point, not a conclusion. Once you have it, work through the responses.
Start with the detractors. Read their comments and group the reasons. Recurring themes point to specific problems you can act on, such as slow support or a confusing onboarding step. Fixing the top one or two reasons is where most gains come from.
Then look at your passives. They scored 7 or 8, so they are close to promoting you but held back by something. Their comments often reveal a small gap between good and genuinely loyal.
Track the score over time, not as a one-off. Field it on a fixed cadence, keep the method identical, and watch the direction. A score moving from 25 to 32 over three quarters is a clearer signal of progress than any single reading.
Finally, connect the score to action. Assign the top detractor themes to owners, make a change, and check whether the next wave moves. NPS is most useful as a loop: measure, read the reasons, fix, and measure again.
Build a study, reach verified respondents in your market, and get answers back in days.
Everything a research team does. Without the research team.
With SegmentOS you can build the study, reach a verified audience, and get data you can actually trust. End to end, no research background required.
Data quality
Every study runs through device fingerprinting, speeding detection, attention checks, and screener disqualification.
Start Measuring Customer Satisfaction With NPS
You now have the full method: the question, the 0 to 10 scale, the formula, the sample sizes, and the benchmarks. The last piece is fielding it to respondents you can trust.
SegmentOS ships CSAT and NPS templates with the standard questions, screeners, and attention checks already set up. Every study runs through six layers of quality checks, so responses that fail never reach your dataset. You launch the study and read a score you can defend.
Frequently Asked Questions (FAQ)
What is a good NPS score?
Above 0 is good, above 20 favorable, above 50 excellent, and above 80 world-class, per Bain's framework as reported by Qualtrics. Judge yours against your industry and your own trend.
How do I calculate NPS?
Subtract the percentage of detractors (scores 0 through 6) from the percentage of promoters (scores 9 or 10). Passives (7 or 8) are excluded. The range is −100 to +100.
How is NPS different from CSAT?
NPS is relational and measures likelihood to recommend on a 0 to 10 scale. CSAT is transactional and measures satisfaction with a specific interaction on a 1 to 5 scale.
How often should I run NPS?
Field it periodically on a fixed cadence, such as quarterly. NPS is not tied to a single interaction, so consistency across waves matters most.
Who invented NPS?
Fred Reichheld, with Bain & Company, in a December 2003 Harvard Business Review article, "The One Number You Need to Grow."
