NPS Calculator: Work Out Your Net Promoter Score and What It Means

Net Promoter Score is calculated by subtracting the percentage of detractors (people who scored you 0 to 6) from the percentage of promoters (9s and 10s). The result is a whole number from -100 to +100. The free Net Promoter Score calculator below does the math, shows how much to trust the result, and compares it against current industry benchmarks. It loads with a worked example; replace the numbers with your own.

Everything runs in your browser. Nothing you type is sent anywhere; the share link just encodes your counts in the URL, and the downloadable score card is drawn on your device.

How is NPS calculated?

The formula is: NPS = % promoters − % detractors. To calculate your NPS, sort every respondent by their answer to one question, "How likely are you to recommend us to a friend or colleague?", asked on a 0 to 10 scale. Each response lands in one of three segments:

Worked example: you collect 250 responses. 140 people score you 9 or 10 (56%), 70 score 7 or 8 (28%), and 40 score 0 to 6 (16%). Your NPS is 56 minus 16: an NPS of 40. By convention it is written as a whole number, not a percentage.

Passives matter more than the formula suggests. They don't appear in the subtraction, but they sit in the denominator, so every passive dilutes your promoter percentage. Moving a passive to a promoter lifts the score exactly as much as moving a detractor to a passive; converting a detractor all the way to a promoter is worth double either single step.

The metric comes from Fred Reichheld's "The One Number You Need to Grow" (Harvard Business Review, December 2003), which proposed the recommend question as a proxy for growth.

What is a good NPS score?

Anything above 0 means you have more promoters than detractors, which is the baseline for "good." The interpretation bands most teams use are the ones Qualtrics attributes to Bain & Company, the firm behind NPS:

Retently's 2026 benchmark analysis draws similar lines from its survey data: 0 to 30 is a good range, above 30 means you're doing great, and over 70 means your customers love you.

Treat the bands as a sanity check, not a grade. Bain's own benchmarking product, NPS Prism, puts it plainly: a positive, higher score is better, but performance depends on several critical variables, starting with your industry. A 42 would be below average for a consulting firm and comfortably above average for a cloud provider.

NPS benchmarks by industry (2026)

The most recent public benchmark data comes from Retently's 2026 report (published April 2026, drawn from a sample of at least 10,000 surveys). Industry averages ranged from 26 to 68:

How many responses do you need for a reliable NPS?

Roughly 300 to 400 responses gets you within about ±10 points at 95% confidence, per the sample-size tables in MeasuringU's analysis of NPS confidence intervals. Below a hundred, the score swings hard: Genroe's worked examples (accessed July 2026) put the point where uncertainty starts to meaningfully shrink at about 100 responses. Run Genroe's example distributions, an NPS of 10 at three sample sizes, through the formula this post uses and the swing is stark (Genroe's own published figures land a point or two higher, using a slightly more conservative two-standard-error convention):

Notice the shape of those numbers: every tenfold jump in responses cuts the margin of error to roughly a third, from about ±51 points at 10 responses to ±16 at 100 and ±5 at 1,000. Precision scales with the square root of sample size, so quadrupling your responses halves your margin of error.

That is why the calculator above reports a margin of error next to your score. An NPS of 40 from 30 responses and an NPS of 40 from 3,000 responses are different facts: the first says "somewhere between roughly 10 and 70," the second pins it down. A quarter-over-quarter move smaller than your margin of error is noise, not news.

Why our margin of error looks bigger than other calculators'

Because it uses the right formula. NPS is a difference of two proportions, and its per-response variance is the promoter share plus the detractor share, minus the square of the score (all as fractions of total responses). In symbols: margin of error = 1.96 × √((promoter share + detractor share − NPS²) ÷ responses), expressed back in points. Online calculators that show a margin of error at all typically borrow the generic single-proportion formula from opinion polling, which can understate NPS uncertainty by up to a factor of two. Wider and true beats narrow and wrong: if this calculator tells you ±14 where another says ±7, trust the ±14.

Response counts are also getting harder to earn. Retently's 2026 response-rate study (25 million invitations across 600 ecommerce brands) measured average NPS survey response rates at just 4.5%, with email surveys overall averaging about 3%, against 32% for in-app surveys. At those rates, banking 100 responses over email means sending a few thousand invitations; an on-site or in-app survey gets there dramatically faster.

How do you improve your NPS?

The score moves when you act on what drives it, and the verbatim comments, not the number, tell you what that is. The reliable pattern: close the loop with every detractor quickly, mine the comments for the two or three recurring complaints, fix the top one, and re-measure. This is what Bain's Net Promoter System calls the inner loop: individual feedback routed straight back, in the customer's own words, so someone closes it with that customer while it is fresh. Systemic problems that show up across many responses belong to the outer loop instead: process and policy fixes, not individual callbacks. The calculator above keys its suggested next step to the same distinction. Chasing the number directly (survey timing tricks, begging for 10s) inflates the score without changing the underlying loyalty, and the gap eventually shows up in churn.

The other half is doing something with your promoters. A 9 or 10 is a raised hand, and most teams let it expire in a dashboard. Asking a promoter for a review, a referral, or a short interview while the goodwill is fresh converts the sentiment into an asset you can put in front of the next buyer. That handoff, from measuring advocacy to using it, is where a score program starts paying for itself; our guide on how NPS fits inside a voice-of-customer program maps that boundary, and Beyond NPS makes the case for measuring advocacy as behavior, not just sentiment.

How do you collect NPS automatically?

Every number in this calculator started as a response somebody had to collect, and collection is the part nobody should be doing by hand. Below is the survey widget Shine puts on your website, live and clickable.

<script async src="https://backend.shine.studio/nps/widget/embed.js?key=pk_live_××××××××××××××××"></script>

Each response lands in your dashboard the moment it arrives, already scored, and trends get a low-confidence flag while the sample is still small.

Where Shine fits (and where it doesn't)

Shine runs NPS and CSAT surveys free, over email and an embeddable website widget, with no cap on responses. The product's actual job begins after the score: when a promoter turns up, Shine invites them into a short AI-conducted interview and turns that one conversation into verified quotes, case studies, and review drafts. If your bottleneck is what to do with happy customers once you find them, that is what it is built for.

To be straight about the boundaries: if you need enterprise survey analytics, think text analytics across millions of responses, complex routing, and statistical tooling, a Qualtrics-class platform is the right category, and our rundown of NPS survey tools compares the like-for-like options honestly. And if all you need today is the math, the calculator above requires no signup and never will.

Frequently asked questions

Can NPS be negative? Yes. The scale runs from -100 (every respondent is a detractor) to +100 (every respondent is a promoter). How alarming a negative score is depends on the company you keep: Bain's NPS Prism benchmarks show utilities averaging just 15 and internet providers 16, so a score five points below zero in those industries trails the pack by roughly twenty points, while the same score in consulting, where Retently's averages run in the high 60s, would signal a full-blown crisis. Distance from your industry's average matters more than the sign.

Why is a 7 or 8 not counted as a positive response? By design, the scale punishes lukewarm sentiment. Reichheld's original research found that people who scored 7 or 8 behaved differently from 9s and 10s: they repurchased and referred at meaningfully lower rates and defected to competitors more easily. Counting them as positive inflated the metric's correlation with growth, so the formula deliberately ignores them.

What's the difference between NPS and CSAT? NPS asks one relationship question (would you recommend us?) and produces a -100 to +100 score; CSAT asks about satisfaction with a specific interaction (this purchase, this support ticket) and is usually reported as the percentage of satisfied responses. NPS tracks the health of the whole relationship; CSAT localizes problems to a touchpoint. Most programs run both: CSAT to find friction, NPS to watch the relationship trend. If you need the CSAT math too, our free CSAT calculator works the same way as this one.

How often should you measure NPS? Relationship NPS is typically surveyed quarterly or twice a year, cycling through your customer base rather than blasting everyone at once, so fatigue stays low and the trend line stays continuous. Transactional NPS is triggered by events (onboarding complete, renewal, support resolution) and follows the event's rhythm. Whatever the cadence, keep methodology constant; changing the channel or timing mid-stream breaks your own trend line, which is the one benchmark that matters most.