Net Promoter Score (NPS) is the most widely used customer-loyalty metric in the world. It is one question — but running it badly leads to bad decisions. This article walks through how to compute it, how to operate it, and how to turn it into action.
What NPS is
NPS was introduced by Fred Reichheld of Bain & Company in 2003. It is a single question:
"How likely are you to recommend our product/service to a friend or colleague?"
Responses are captured on an 11-point scale from 0 to 10. That simplicity is its greatest strength.
How NPS is calculated
Respondents are bucketed into three groups by score, and the score is computed from the ratio of those buckets.
| Group | Score | Description |
|---|---|---|
| Promoters | 9–10 | Enthusiastic fans. Drive growth via repeat and word of mouth |
| Passives | 7–8 | Satisfied but switchable |
| Detractors | 0–6 | Dissatisfied. Source of negative word of mouth |
The formula:
NPS = % Promoters − % Detractors
Passives are deliberately excluded. The score range is −100 to +100.
Example
Out of 100 respondents: 40 promoters, 35 passives, 25 detractors.
NPS = 40% − 25% = +15
Realistic benchmarks
"What is a good NPS?" depends heavily on industry, country, and methodology. As rough guidance:
- Below 0: Many dissatisfied customers. Urgent intervention needed
- 0–30: Industry average
- 30–50: Strong. Many loyal customers
- Above 50: Exceptional. The level of Apple, Costco, and similar world-class brands
That said, change from your last round is more useful than the absolute number.
Why NPS is useful
1. Simple and trackable over time
One question means low respondent burden, so it fits a recurring measurement cadence.
2. Correlated with business outcomes
Multiple studies show higher-NPS firms have higher revenue growth and retention (the strict causality is debated, but the correlation is reliable enough to operate on).
3. Easy to communicate up
A single number is easy to put in board reports and exec dashboards.
Common misreadings
"The score itself is the goal"
"Lift NPS by 5 points" sounds like a target — and it is a trap. Understanding why detractors are detractors produces more long-term value than chasing the score directly.
"High score = we're fine"
A high NPS with low response count or biased sampling is not trustworthy. Always check who actually responded.
"We should compare to industry average"
Industry averages move with method (email vs. in-store, US vs. EU vs. APAC). Trend against your own history is more decision-useful.
What to ask alongside NPS
The number alone does not tell you why. The standard set of three:
- The NPS question — 0 to 10
- Reason (open text) — "Tell us why you scored it that way"
- What to improve (open text, optional) — "What would you like us to change?"
These three questions give you the quant (NPS) and the qual (reasons) together. Every extra question lowers your response rate, so keep it to these three by default. If you also need attributes (plan, tenure, role), add only one or two and aim for five questions total to limit drop-off. Making the reason field required rather than optional dramatically improves the yield of your downstream open-text analysis.
How to act on it
Step 1: Follow up with detractors
Read every open-text response from detractors (0–6). Where possible, contact them individually. Just "Thanks for the feedback — can we hear more about it?" is enough to surface churn signals and repair relationships.
Step 2: Activate promoters
Promoters (9–10) are the natural audience for review requests, referral programs, and case-study interviews.
Step 3: Cluster the open-text answers
Once you have 10–20+ open responses, classify them into themes. Patterns like "support response time," "feature friction," and "pricing" usually emerge.
Cadence
- Relational NPS: every 6–12 months, sent to all customers. Tracks long-term trend
- Transactional NPS: fired right after an event (purchase, support case). Used for measuring specific initiatives
Ideal is to run both, but starting with annual relational NPS is a perfectly viable first step.
As a practical rule, leave at least three months between relational rounds. Shorter intervals trigger survey fatigue ("not this again"), which drags down response rate and skews the sample. Transactional NPS is the opposite: send it right after the event — ideally within 24–48 hours — or memory fades and answer quality drops.
NPS in B2B
B2B (business-to-business) NPS works on very different assumptions than B2C. The biggest difference: a single customer account contains multiple stakeholders, and willingness to recommend varies sharply by role.
- Decision-makers / buyers: tend to judge on ROI and business impact. Their score tracks closely with commercial outcomes.
- End users on the ground: judge on day-to-day usability. It is not unusual for their score to land at the opposite end from the buyer's.
- Admins / operations: weight onboarding effort, hand-off, and support quality.
So if you do not design who you ask, the same account can produce wildly different scores. The rule of thumb is to capture role and usage frequency as respondent attributes and segment the results accordingly. Because B2B accounts are few but high-value, it also pays to wire detractor follow-up into account management — a single detractor can represent a large slice of revenue.
For concrete question design (asking by role, ARR-weighted aggregation, extra questions for buyers), see B2B NPS question design.
Summary
NPS combines simplicity and business signal in a way few metrics do. The three things that matter:
- Do not get whip-sawed by score movements
- Always pair the number with open-text
- Follow up with detractors — convert insight into action
Repoan ships an NPS standard template and a B2B NPS template, both one-click to deploy. As responses come in, the AI report feature (see AI response analysis) classifies open text into themes like "pricing / support / features" automatically, so promoter and detractor patterns surface immediately.
When to use NPS vs. CSAT is covered in NPS vs CSAT.
Related
- Employee version: eNPS — measuring and improving employee NPS
- B2B-specific design: B2B NPS question design
- After collection: Putting survey results to work
- Continuous operation: Running a survey PDCA cycle
- AI-era data strategy: Why run surveys in the AI era