Quick definition: Net Promoter Score (NPS) is a survey-based loyalty indicator calculated as the percentage of respondents who are promoters minus the percentage who are detractors after answering how likely they are to recommend a product or company on a zero-to-ten scale.
What is Net Promoter Score?
NPS asks a standardized question, commonly “How likely are you to recommend us to a friend or colleague?” Respondents who select 9 or 10 are classified as promoters, 7 or 8 as passives, and 0 through 6 as detractors. The score ranges from -100 to 100. It is a directional measure of stated willingness to recommend, not a direct count of referrals, satisfaction, revenue, retention, or product quality.
The calculation is NPS = percentage of promoters − percentage of detractors. If 50% of 200 respondents are promoters and 20% are detractors, NPS is 30. Passives affect the denominator but contribute neither positively nor negatively. Always report response count and category distribution with the score: two groups can share an NPS while having very different numbers of promoters and detractors.
Teams use NPS as one signal in a voice-of-customer program. It can identify themes, monitor a stable population over time, and prompt follow-up research. It should sit alongside behavioral engagement, retention, support, reliability, and task-success measures. A person may recommend a brand yet struggle with a specific product workflow; another may use a product successfully but rarely recommend any work tool.
Designing an NPS measurement program
Specify who receives the survey, when they receive it, how often, the channel, language, question wording, and treatment of incomplete responses. Relationship NPS measures a broader ongoing perception, often sampled periodically from customers. Transactional NPS follows a discrete interaction, such as support resolution or onboarding completion. Do not blend them into one trend: they answer different questions and have different response biases.
| Design choice | What to document | Example |
|---|---|---|
| Population | Eligible customer role and exclusions | Active paid workspace administrators. |
| Cadence | Frequency cap and trigger timing | Quarterly, no more than once per 180 days. |
| Mode | Email, in-product, phone, or other channel | In-product after a completed workflow. |
| Question | Exact wording, scale, locale | Standard 0–10 recommendation question. |
| Follow-up | Open-text prompt and response handling | “What is the main reason for your score?” |
Use a stable survey design for trend interpretation. Moving from email to an in-product prompt, changing copy, or sampling only highly active customers can shift response propensity and the score without changing sentiment. If a redesign is necessary, run both approaches in parallel for a period or annotate the structural break. Keep a denominator of invitations sent and calculate response rate by segment.
Open-text responses provide context but need disciplined coding. Establish a codebook, allow multiple themes, record unknown or mixed feedback, and sample-review coding quality. Do not calculate precise percentage changes from a handful of memorable comments. Link themes to product areas and operational owners, then investigate with usage and support data.
Interpreting NPS responsibly
Compare like with like: the same population, cadence, channel, locale, and product stage. External benchmarks are rarely directly comparable because samples, industries, survey modes, and response rates differ. A score can be useful even without a benchmark if it is measured consistently and leads to a better understanding of customer outcomes.
Survey estimates have uncertainty. A score from 40 respondents is less stable than one from 4,000, and nonresponse can matter more than sampling variation. Report invitation count, completed responses, response rate, promoter/passive/detractor shares, and a confidence interval or other uncertainty statement where decisions depend on small differences. Segment results only when each segment has sufficient data and a concrete action attached.
Follow up with care. Detractor outreach can help resolve a problem, but it should not pressure customers to revise scores or selectively suppress feedback. Close the loop by acknowledging themes, prioritizing changes, and explaining what was learned. The aim is improving the experience, not increasing a dashboard number.
Experiment scenario: onboarding guidance
A B2B product receives recurring NPS comments that new administrators feel uncertain during data setup. The team hypothesizes that an onboarding guide with clear permission explanations and a progress checklist will reduce effort and improve early customer confidence. It does not make NPS the sole primary metric because an early survey response is sparse and can be influenced by the new prompt itself.
Eligible new administrator accounts are randomized before onboarding. The primary outcome is completed, error-free first data connection within 14 days. Secondary outcomes include time to first report, support contacts, and a pre-planned transactional NPS survey sent at the same point and through the same channel for both variants. Guardrails include security warnings, permission escalation, survey response rate, cancelations, and negative open-text themes about pressure or confusion.
The analysis includes all assigned accounts. It does not compare only people who answer the survey, since response can be affected by treatment and differs by satisfaction. If task completion rises while NPS remains unchanged, the guide may solve a functional barrier without changing recommendation intent. If NPS rises but completion and support do not, examine survey placement and response bias before claiming the experience improved.
Interpretation and data limitations
NPS measures a stated intention in a sampled respondent population. People who answer surveys can differ from people who ignore them in tenure, enthusiasm, frustration, language, and access. In-product surveys can miss inactive customers; email surveys can miss people with filtered or outdated addresses. A higher response rate does not automatically remove bias, and a lower rate does not automatically invalidate every result.
Scores are sensitive to context. A recent outage, renewal negotiation, support interaction, seasonal workload, or public news can affect answers. For multi-seat accounts, an administrator and daily end user may provide very different perspectives. Aggregate account-level decisions cautiously and keep respondent role visible. Privacy rules and consent obligations may limit contact and follow-up; honor those limits.
Most importantly, correlation with retention or growth does not make NPS a causal driver. Promoters may be more likely to remain because they already receive more value. Treat NPS as diagnostic evidence, and test specific experience changes using behavioral outcomes and appropriate experimental design.
Common mistakes
- Using NPS as a complete health metric: pair it with product, support, and retention outcomes.
- Changing survey design mid-trend: preserve comparability or annotate the break.
- Reporting only the score: include respondents, response rate, and category shares.
- Surveying too often: cap frequency and respect customer attention.
- Pressuring customers for high ratings: protect independent feedback and trust.
- Calling score movement causal: investigate bias and test concrete product changes.
FAQ
How is NPS calculated?
Subtract the percentage of respondents scoring 0–6 from the percentage scoring 9–10. Scores of 7–8 remain in the denominator but do not add or subtract.
What is a good NPS?
There is no universal threshold. Compare a stable, relevant population over time and focus on actionable feedback and customer outcomes.
Is NPS the same as customer satisfaction?
No. NPS asks about willingness to recommend; customer satisfaction surveys usually ask about satisfaction with a product or interaction.
Should NPS be the primary experiment metric?
Rarely by itself. It is often sparse and noisy. Use a direct task or value outcome as primary and NPS as a complementary customer signal.
Summary
NPS is the percentage of promoters minus detractors on a standardized recommendation question. Its usefulness depends on a stable survey design, clear population, transparent response data, and thoughtful qualitative follow-up. Interpret it with uncertainty and behavioral outcomes, and test specific improvements rather than optimizing the score alone.