Quick answer: Net Promoter Score (NPS) is calculated with one formula: NPS = % of Promoters − % of Detractors. Ask customers "How likely are you to recommend us?" on a 0–10 scale. Promoters score 9–10, Passives score 7–8, Detractors score 0–6. Subtract the percentage of Detractors from the percentage of Promoters (ignore Passives) to get a number between −100 and +100. That is the entire calculation. The hard part — and the part that actually grows the business — is understanding why customers gave the scores they did. This guide covers the formula, a worked example, 2026 benchmarks, and how to capture the reasons behind every score automatically.
The NPS formula, step by step
NPS comes from a single survey question — "On a scale of 0 to 10, how likely are you to recommend [product] to a friend or colleague?" — and three response buckets:
| Bucket | Score range | Meaning |
|---|---|---|
| Promoters | 9–10 | Loyal enthusiasts who fuel growth and referrals |
| Passives | 7–8 | Satisfied but unenthusiastic; easily swayed by competitors |
| Detractors | 0–6 | Unhappy customers who can damage your brand via word of mouth |
To calculate NPS:
- Collect responses to the 0–10 recommendation question.
- Categorize each response into Promoter, Passive, or Detractor.
- Convert each group to a percentage of total responses.
- Subtract: NPS = % Promoters − % Detractors.
Passives count toward your total response number (the denominator) but are not added or subtracted directly — they quietly drag your score down by diluting the promoter percentage.
A worked example
Say you survey 200 customers and get:
- 120 Promoters (scored 9–10) → 120 ÷ 200 = 60%
- 50 Passives (scored 7–8) → 25% (used in the denominator, not added)
- 30 Detractors (scored 0–6) → 30 ÷ 200 = 15%
NPS = 60% − 15% = +45.
Your score is 45, not 45%. NPS is always expressed as a whole number between −100 (every respondent is a Detractor) and +100 (every respondent is a Promoter). A useful sanity check: if every customer scored 9–10, you would have an NPS of 100; if every customer scored 0–6, you would have −100.
What is a "good" NPS in 2026?
A score is only meaningful against a benchmark. Based on 2026 data across 150,000+ organizations:
- The average NPS is around 32, with a median near 44.
- Above 30 is strong, above 50 is excellent, and above 70 puts you in the global top tier. The top 25% of companies score 72 or higher; the bottom 25% score 0 or lower.
- Industry matters enormously. B2C companies outperform B2B by roughly 11 points (about 49 vs. 38). SaaS and software cluster near the bottom — B2B SaaS averages around 36, so 40+ is a strong target — while sectors like manufacturing can post medians around 65.
The practical takeaway: always benchmark against your own industry, not a universal number. A 35 can be excellent in one vertical and mediocre in another. Track your trend over time — your own quarter-over-quarter movement is often more actionable than any external benchmark.
The limit of the number: NPS tells you what, not why
Here is the problem every team eventually hits. NPS is a fantastic thermometer and a useless diagnosis. It tells you a customer is a Detractor; it does not tell you whether they are leaving because of price, a missing feature, a support failure, or onboarding friction. A score with no reason behind it cannot be acted on — and the classic follow-up ("What's the primary reason for your score?") gets one-line answers that are too thin to code or trust.
This is exactly where most NPS programs stall. Teams accumulate a trend line but can't explain the movement, so the metric becomes a vanity number reported in a dashboard rather than a driver of decisions. (We have written before about why NPS is broken and the best NPS alternatives when the single number stops earning its place.)
How to capture the "why" behind every score with AI
The fix is to treat the NPS score as the opening of a conversation, not the end of a survey. Instead of a static "reason" textbox, an AI moderator can ask the score question and then probe in real time: a Detractor who mentions "it's too expensive" gets asked what they expected to pay and what would make it worth it; a Promoter who says "the reports save me hours" gets asked which report and what they did before. That follow-up depth is the difference between a number and an insight.
With Koji, this is native:
- Use the scale question type to ask the 0–10 NPS question, so the quantitative score is captured cleanly and visualized as a distribution. See structured questions — Koji supports six types (open-ended, scale, single-choice, multiple-choice, ranking, and yes/no) in a single study.
- The AI moderator then asks dynamic, open-ended follow-ups based on the score and the reason given — the probing a one-line survey box never does. See AI-moderated interviews and voice vs text interviews.
- Every transcript is auto-coded, so the reasons behind Detractor and Promoter scores cluster into themes automatically across hundreds of responses. See AI auto-tagging and the thematic analysis guide.
- Then chat with your transcripts — ask "what's the top reason Detractors gave?" and get a cited answer in seconds.
The result: your NPS number and the ranked list of reasons behind it, in one study, without manually reading and coding open-ended replies — a task that traditionally costs researchers four to seven hours of work per hour of interview audio.
Common NPS calculation mistakes to avoid
- Counting Passives as Promoters. Passives (7–8) are not added to your score; treating them as positives inflates the number.
- Reporting NPS as a percentage. It is a whole number from −100 to +100, not a percent.
- Tiny samples. A handful of responses produces a volatile score; collect enough responses for the percentages to stabilize before you trust the number.
- Surveying only happy users. Sampling bias (e.g., only emailing recently active customers) inflates NPS. Survey a representative cross-section.
- Stopping at the number. The score without the reason cannot drive a roadmap. Always capture the "why."
From score to action
NPS earns its keep only when it changes what you build. The workflow that works in 2026: measure the score with a clean scale question, capture the reason with AI-moderated follow-up, let thematic analysis rank the drivers, then route the top Detractor themes to the teams that own them — and re-measure next quarter. That closes the loop between a metric and a decision. For the broader system around this, see our guides to the customer feedback loop and churn survey questions.
The bottom line
Calculating NPS is genuinely simple: % Promoters − % Detractors, on a 0–10 scale, expressed as a number from −100 to +100. Benchmark it against your own industry — roughly 30+ is strong, 50+ excellent, with B2B SaaS averaging around 36 — and track your own trend over time. But the calculation is the easy 10%. The 90% that grows the business is understanding why customers scored the way they did, and that requires real conversation, not a one-line textbox.
Want the score and the story behind it? Start with Koji free — run an AI-moderated NPS study with a clean scale question plus dynamic follow-ups, and get auto-coded reasons behind every score in hours. 10 credits, no credit card, no research expertise required.
