# NPS: Calculate the Score Without Losing the Customer’s Voice NPS calculation, interpretation, limitations, and the right open-ended follow-up: connect the score to causes, evidence, and decisions. - Canonical URL: https://www.harmate.com/en/blog/nps-calculate-the-score-without-losing-the-customers-voice - Author: Harmate Team - Published: 2026-08-20 - Updated: 2026-08-20T23:36:59.479945+00:00 - Language: en ## Content # NPS: Calculate the Score Without Losing the Customer’s Voice The Net Promoter Score, or NPS, fits into one question and one subtraction. That simplicity explains its popularity: a number is easy to track, share, and compare over time. It also creates the metric’s main risk. Once the score becomes the goal, it can replace the conversation it was meant to start. An NPS of 32 describes the balance between promoters and detractors. It does not explain what changed, why two customers gave the same rating for opposite reasons, or which action deserves priority. For that, you need context and the respondents’ own words. This guide explains how to calculate NPS correctly, interpret it cautiously, and complete it with an open-ended question that turns a metric into material for a decision. ## What is NPS? NPS measures a stated intention to recommend. The classic wording asks: > On a scale from 0 to 10, how likely are you to recommend [the company, product, or service] to a friend or colleague? Responses are placed into three groups: - **promoters**: ratings of 9 or 10; - **passives**: ratings of 7 or 8; - **detractors**: ratings from 0 to 6. This classification creates a summary indicator. It does not automatically turn promoters into advocates, passives into customers ready to leave, or detractors into opponents. These are calculation categories, not psychological profiles. ## How do you calculate NPS? The formula is: **NPS = percentage of promoters − percentage of detractors** Passives remain part of the total response count but are neither added nor subtracted. The final score therefore ranges from -100 to +100. Consider 200 valid responses: - 92 promoters, or 46%; - 68 passives, or 34%; - 40 detractors, or 20%. The NPS is **46 − 20 = 26**. Three precautions prevent misleading calculations: 1. calculate percentages using valid answers to the NPS question; 2. keep the wording, scale, and collection mode comparable when tracking change; 3. display the response count next to the score. An NPS of 40 based on 15 answers does not carry the same descriptive weight as the same score based on 1,500 answers. AAPOR recommends keeping question wording, framing, and methodology as similar as possible when measuring change over time. A change in questionnaire or channel can create an apparent shift even when the underlying experience has not changed. ## What is a good NPS? The short answer is frustrating but useful: **a good NPS is a score that can be interpreted in context**. A positive score only means the share of promoters is larger than the share of detractors. It does not prove that the experience is good, much less that the score will cause growth. Expectations, rating cultures, industry, relationship type, and survey timing all affect responses. Three comparisons are usually stronger than a universal benchmark: - the same population over time under a stable protocol; - comparable segments measured at the same time; - specific stages of the same journey, when the question is asked under consistent conditions. Academic research does not support treating NPS as the best predictor of growth in every setting. A *Journal of Business Research* study using more than 193,000 responses found that several recommendation-based methods were strongly correlated and that some alternatives predicted growth better in the dataset studied. A 2026 longitudinal study also emphasizes that predictive validity depends on context. NPS can be a management signal; it should not be treated as causal proof. ## The NPS question is not enough Two customers may both give a 6. One faced an unusual delivery failure. The other believes the product no longer fits their need. The score puts them in the same category, while the decisions differ: repair a process in one case and revisit the value proposition in the other. The reverse is also true. Two promoters may give a 10 because one values support and the other values price. A higher NPS does not reveal the mechanism that should be protected. This is where the open-ended question matters. Closed responses make measurement and comparison easier. Open responses are richer for discovering unexpected causes, contradictions, exceptions, and the language customers actually use. The two formats are not rivals: the score tells you **where to look**; the verbatim answer helps explain **what is happening**. ## Which open-ended question should follow NPS? The most robust wording is short and neutral: > What is the main reason for your rating? It works for every score and avoids assuming that a low rating necessarily means disappointment or that a high rating necessarily means enthusiasm. Other wordings may fit a specific decision: - **To understand a moment**: Which part of your experience influenced your rating the most? - **To identify a change**: What should change first to improve your experience? - **To protect a strength**: What matters most in your willingness to recommend us? - **To collect concrete evidence**: Could you describe the situation that most influenced your rating? Avoid leading wording such as Why are you dissatisfied? after a low rating. It imposes an interpretation before the person has spoken. Also avoid asking for the cause, the solution, and a competitor comparison in the same field. Double- or triple-purpose questions tend to produce answers that are difficult to interpret. One well-placed open field is often better than a series of mandatory text boxes. It limits burden, keeps the survey short, and leaves room for what you did not anticipate. ## When should you ask the question? NPS is commonly used in two ways. **Relational NPS** seeks an overall view of the relationship. It can be collected at regular intervals from a consistent sample. It supports trend monitoring, but the cause of a change may be diffuse. **Transactional NPS** follows an identifiable interaction: a purchase, delivery, support exchange, or onboarding step. Its context is more precise, but a single episode should not be generalized to the whole relationship. In both cases, choose timing from the decision you expect to make. A survey sent immediately after an order rarely measures actual product use. One sent long after an incident relies more heavily on memory. Record the trigger, delay, and surveyed population so waves remain comparable. ## How should you analyze NPS comments? Reading only detractor comments creates an incomplete picture. Passives may reveal the frictions that keep a correct experience from becoming remarkable. Promoters identify qualities worth protecting. A useful analysis can follow six steps. ### 1. Keep the rating linked to the response Every comment should remain connected to its rating, collection moment, and authorized segments. Do not detach the text from its provenance. ### 2. Read a sample before creating categories Do not begin only with a list of reasons chosen internally. An open first reading can reveal topics that predefined categories would have missed. ### 3. Group causes, not just words Delivery and parcel may describe the same cause. Conversely, fast can praise support or criticize a rushed decision. Context matters more than raw word frequency. ### 4. Compare themes across NPS groups For each theme, inspect its presence among promoters, passives, and detractors. A topic that appears everywhere may be structural without explaining the score gap. A less frequent theme concentrated among detractors may reveal a specific failure. ### 5. Return to the evidence A theme is not a self-sufficient conclusion. Keep several source comments, including evidence that contradicts the dominant interpretation. Decision-makers must be able to verify the meaning assigned to a group. ### 6. Connect each insight to a decision Assign an owner, a testable action, and a review point. If a theme changes no decision, it may be interesting but not yet a priority. ## Mistakes that make NPS less trustworthy ### Comparing different populations A score from customers who contacted support is not directly comparable with a score from the whole customer base. Always state who was invited and who responded. ### Confusing variation with cause An increase after a feature release does not prove that the feature caused it. Other changes, seasonality, or respondent composition may be involved. ### Managing teams only to the score When compensation depends directly on NPS, teams may select respondents, pressure them about the rating, or avoid certain audiences. The number gets cleaner while the signal gets worse. ### Hiding passives They do not directly affect the formula, but they may represent the largest response group. Ignoring their explanations removes an important space for improvement. ### Summarizing comments without sources A synthesis without access to original responses can smooth out contradictions or amplify an anecdote. Every important theme should remain verifiable. ## A genuinely actionable NPS protocol Before launching the survey, write down these six elements: - **decision**: what you can change after seeing the results; - **population**: who is invited and under which rule; - **moment**: the event or frequency that triggers collection; - **closed question**: stable NPS wording; - **open question**: a short, neutral request for the main reason; - **expected evidence**: themes, source comments, segments, and the related action. This framework prevents the collection of a score with no owner or next step. It also makes limitations explicit: nonresponse, segment size, protocol changes, and unconfirmed hypotheses. ## Where Harmate can help, quietly Harmate does not replace your definition of NPS or your responsibility for interpretation. It can help with the part the score compresses: designing a questionnaire that preserves real signal potential, collecting open responses, organizing insights, and returning to the verbatim evidence that supports them. The questionnaire’s Initial Check examines clarity, burden, and qualitative angles before fieldwork. After collection, thematic analysis organizes concepts while preserving source evidence. The final decision remains human. ## Key takeaway NPS is useful when it stays in its place: a simple indicator, comparable under a stable protocol, and modest enough to start an investigation rather than end the discussion. Calculate it correctly. Show the response count. Compare comparable populations. Then ask a neutral open-ended question, analyze causes in context, and keep the source comments accessible. The score tells you that there is a gap. Your customers’ words help you understand which one — and decide what to do. ## Sources - [AAPOR — Best Practices for Survey Research](https://aapor.org/standards-and-ethics/best-practices/) - [Bain & Company — Assessing Your Net Promoter System](https://www.bain.com/contentassets/5df724cb5e144d2ca6ebfafbdd71cd79/loyalty_insights_assessing_your_net_promoter_system.pdf) - [Baehre et al. — Customer mindset metrics: a systematic evaluation of NPS vs. alternative calculation methods](https://doi.org/10.1016/j.jbusres.2022.04.048) - [Matzler et al. — Testing the predictive validity of the Net Promoter Score in ski resorts](https://doi.org/10.1007/s11846-026-00986-2)