The standard NPS question is fixed. The follow-up does all the work. Six follow-up variants, branched logic, and the pitfalls to avoid.
"On a scale of 0 to 10, how likely are you to recommend us?" is fixed — the wording has been the standard since the early 2000s, and changing it breaks comparability across the entire NPS literature. The question that varies, the question that produces the verbatim, the question almost no team thinks about carefully — that is the follow-up. Sopact reads every follow-up verbatim on arrival, against the same contact's prior wave, so the follow-up question becomes the most load-bearing piece of survey design the team will write.
By Unmesh Sheth · Founder & CEO, Sopact · Updated May 2026
An NPS survey is two questions. The standard rating question on a 0-to-10 scale, and an open-ended follow-up. That is the entire instrument. Surveys that bolt on five demographic questions, four product questions, and three free-text fields are not NPS surveys — they are general feedback forms with NPS at the top. The narrowness is the point.
"On a scale of 0 to 10, how likely are you to recommend [company/product/service] to a friend or colleague?" Fixed wording, fixed scale. The whole point is comparability across the literature and across the team's own waves. Changing this question breaks two decades of benchmark data and most internal trend analysis.
"What is the primary reason for your score?" is the most common default. This is where instrument design actually happens. The wording, the framing, whether the question branches by score range — all of these decisions shape whether the verbatim that arrives is actionable or generic.
The standard NPS question is fixed wording on a fixed scale. The scale runs 0 to 10, with three buckets: 0 through 6 are detractors, 7 and 8 are passives, 9 and 10 are promoters. Two decades of benchmark data are built on this exact wording.
"On a scale of 0 to 10, how likely are you to recommend [company] to a friend or colleague?"
0–6 detractors · 7–8 passives · 9–10 promoters
Substitute the company, product, or service name. Do not change the rest of the wording. "How likely" is the standard verb; "to recommend" is the standard intent; "friend or colleague" is the standard recipient set. The wording was set by Fred Reichheld and Bain in the early 2000s — the rules of the game.
The default follow-up is generic. Most NPS surveys ask "what is the primary reason for your score?" because the default template asks it. The default is fine and produces fine verbatims — which is the problem.
"What is the primary reason for your score?"
This is the question that arrives in almost every NPS survey worldwide. It works. It produces a verbatim. And it is leaving almost all of the design value on the table. The follow-up question is where the team gets to shape what kind of verbatim arrives — whether it names a specific fix, names what to protect, names a moment, or names nothing useful at all.
Compare the default to a more designed follow-up. "What would have made this a 10?" elicits specific fixes from detractors and passives. "What should we keep doing?" surfaces what promoters value — the things the team must protect through any product change. "Tell us about a specific moment that drove your score" anchors the verbatim on an actual event, which is far more actionable than an abstract reason.
None of these is universally better. The right follow-up depends on what the team will actually do with the verbatim. The question is: which kind of verbatim does the team need most this quarter? Answer that, and the follow-up writes itself.
For twenty years the survey-design conversation around NPS focused on the rating — whether 0-10 was the right scale, whether the question should be more or less specific, whether to translate "recommend" into industry language. Most of that conversation was settled. The 0-10 wording is the standard, the benchmarks are built on it, the team should not touch it.
The conversation that was not happening was about the follow-up. Teams treated it as boilerplate. They asked "what is the primary reason for your score?" because the template did. They read the resulting verbatims when they had time, summarized them quarterly, and reported the score.
That conversation has to happen now. The reading bottleneck is gone — every follow-up verbatim can be classified on arrival against the team's own codebook. The constraint that justified the default is gone with it. What survives is the design question: what kind of verbatim do we want, and what wording will produce it?
The rating produces the number. The follow-up produces the action. A team that gets the follow-up right gets actionable verbatims every wave. A team that uses the default gets generic verbatims that summarize fine and act on nothing.
This is the same locked argument that anchors /use-case/nps-analysis — expressed here through the instrument-design frame. The pillar covers analysis broadly; this page focuses on the survey question itself.
Six variants of the follow-up, each designed for a different intent. Pick the one that matches what the team will actually do with the verbatim this quarter.
When to useThe fallback. Use this when the team does not yet know what to ask for. It produces verbatims; it does not shape them. Generic but workable. Almost every NPS survey defaults here.
When to useBest for detractors and passives. Anchors the answer on the gap to enthusiasm rather than the reasons for current dissatisfaction. The wording produces sharper, more actionable verbatims than "what is the reason for your score."
When to useBest for promoters. Surfaces what the team must not break through a redesign, a price change, a new release. The promoter verbatims become a protection list, not just a referral pool.
When to useBest for transactional NPS — after a support ticket, a release, a demo. Anchors the verbatim on an actual event rather than an abstract reason, which makes the comment far more actionable.
When to useAction-oriented, focused. Forces a single priority instead of a list of complaints. Works well for transactional NPS and for teams who need to prioritize among many possible fixes.
When to useBest for promoters when the team needs language that can be lifted into testimonials, customer stories, or referral campaigns. The wording prompts response in plain language rather than corporate-speak.
A program can run more than one of these in rotation across waves, or branch the follow-up by score range (see next section). The wrong move is to pick a follow-up once and never revisit it — the question is one of the few NPS variables the team controls.
The same follow-up question is the wrong fit for a detractor and a promoter. Branching the follow-up by score range is the single highest-yield instrument-design move available — three different questions, three distinct kinds of verbatim, one survey instrument.
Detractors named a gap; this question asks them what would close it. Far more actionable than asking why they are unhappy.
Passives are the under-rated segment in NPS — satisfied but not enthusiastic. This question names the gap to enthusiastic recommendation, which is often a single specific thing.
Promoters know what is working. This question turns their answer into a protection list — the team's "do not break this through any redesign" set.
Branched follow-ups make the data slightly harder to aggregate across the response set, because each segment produces a different kind of verbatim. That is a feature, not a bug. The whole point of branching is that detractor verbatims and promoter verbatims should not be averaged together. The team that wants a single aggregate sentiment number does not need this design; the team that wants action does.
Each of these shows up in NPS surveys at least once a quarter. Each is fixable in fifteen minutes. Each costs verbatim quality across every wave it persists.
"How likely are you to recommend us?" becomes "How likely would you be to recommend us in the future?" or "Would you recommend us?" Each wording shift breaks comparability across waves. The rating question is fixed for a reason — leave it alone, design the follow-up instead.
Asking a 10 "what would have made this a 10?" produces an awkward verbatim. Asking a 2 "what should we keep doing?" produces sarcasm. Branching the follow-up by score range (see previous section) is the single most-skipped survey-design move. Three minutes to set up. Years of better verbatim data.
The NPS survey becomes a general feedback form with five demographic questions, four product satisfaction ratings, two pricing questions, and three free-text fields. Response rate falls, data becomes incomparable to benchmarks, and the follow-up verbatim gets diluted by the other questions. The NPS survey is two questions. Keep it two.
Sending NPS right after a support ticket conflates a transactional moment with the relationship. The score crashes after a bad ticket and bounces back the next wave — meaningless variance the team has to explain. Relational NPS goes out on a schedule, not after an event. Use CSAT after the event; use NPS quarterly.
Anonymity feels respectful and produces unactionable data. Without a persistent contact ID across waves, every response is a stranger; the trajectory of a specific customer is invisible; the converted detractor cannot be saved because the team cannot tell which customer they were last quarter. Identified responses, properly handled, produce the strongest signal NPS can give.
This page covers the survey instrument — what to ask and how to ask it. Three adjacent reads cover what to do with the answers: the broader methodology, the verbatim analysis, and the full closed-loop workflow.
What NPS analysis means in 2026 — the methodology, the AI-era thesis, the broader treatment.
Read the pillar →The instrument — the standard rating, the follow-up bank, the branched logic, the pitfalls to avoid.
This pageWhat to do with the follow-up answers — the commercial sub-hub for teams shopping a verbatim-reading tool.
Read the sub-hub →The survey produces the verbatim. The verbatim is what the team acts on. This page designs the question; the rest of the cluster handles the answer.
Same standard rating, three different follow-up questions, three different kinds of verbatim that arrive in the inbox. The follow-up wording is the variable; the verbatim quality is the result.
A SaaS CS team runs relational NPS quarterly with branched follow-ups: detractors get "what would have made this a 10?", passives get "what would move you from a 7 to a 9?", promoters get "what should we keep doing?" Each segment produces actionable verbatims for a different downstream owner.
A cohort training program runs end-of-module NPS with "tell us about a specific moment that drove your score." The verbatim anchors on an actual classroom event — a concept that landed or did not, a facilitator interaction, a peer exchange. The next cohort's curriculum can be redesigned on the named events, not on summary themes.
A scholarship program runs NPS-style feedback six months after award with "what should the program keep doing?" The awardee verbatims become a "what worked, in your own words" appendix for the board report. The next cycle's design protects the things named most often.
Your current wording, your last wave of verbatims. Sixty minutes. No demo accounts.
The standard NPS question is: "On a scale of 0 to 10, how likely are you to recommend [company/product/service] to a friend or colleague?" Respondents who answer 9 or 10 are promoters; 7 or 8 are passives; 0 through 6 are detractors. The wording was set by Fred Reichheld and Bain in the early 2000s and has been the standard ever since. Changing the wording breaks comparability across the literature and across benchmarks.
The follow-up is the open-ended question that comes right after the 0-to-10 rating. The most common default is "What is the primary reason for your score?" or some variation. The follow-up is where the verbatim arrives — the part of the response the team can actually act on. Almost no team thinks carefully about the follow-up question; that is the most consequential survey-design decision in the entire NPS program.
No. The standard 0-10 question is fixed for a reason — comparability across the literature, across benchmarks, across years. Change the recipient name, change the cadence, change the channel — but keep the wording. The room to design is the follow-up question, not the rating question.
Six high-yield variants. "What is the primary reason for your score?" (the default, fine but generic). "What would have made this a 10?" (best for detractors — elicits specific fixes). "What should we keep doing?" (best for promoters — surfaces what to protect). "Tell us about a specific moment that drove your score" (best for transactional NPS — anchors on event). "What is one thing we could improve?" (action-oriented, broad). "What would you tell a friend about us?" (referral-language, useful for testimonial mining). The right choice depends on what the team will actually do with the verbatim.
Yes — branching the follow-up question by score range is one of the highest-yield instrument-design moves available. Detractors get "What would have made this a 10?" — elicits specific actionable feedback. Passives get "What would move you from a 7 to a 9?" — surfaces the gap to enthusiastic recommendation. Promoters get "What should we keep doing?" — protects what is working. Three questions, three response types, three distinct kinds of verbatim — all on the same survey instrument.
Two. The standard 0-10 rating and the open-ended follow-up. That is the entire NPS instrument. Surveys that bolt on five demographic questions, four product questions, and three free-text fields are not NPS surveys; they are general customer surveys with NPS at the top. The strength of NPS is its narrowness; adding questions weakens the response rate without strengthening the signal.
NPS asks "how likely are you to recommend?" on a 0-10 scale. CSAT asks "how satisfied were you with this experience?" usually on a 1-5 or 1-7 scale. NPS measures intent toward the relationship; CSAT measures satisfaction with a transaction. Both have a follow-up open-ended question; both lose the verbatim when teams stop at the score. See /use-case/nps-vs-csat for the full comparison.
Five recurring pitfalls. Generic follow-up wording that produces generic verbatims. Asking the same follow-up question regardless of score (no branching). Adding extra questions to the NPS survey and treating it as a general feedback form. Changing the rating question wording across waves so the data is not comparable. Sending the NPS survey at the wrong moment (right after a support ticket, conflating transaction with relationship).
Relational NPS goes out on a fixed schedule — usually quarterly. The cadence matters less than the consistency: same time of quarter, same channel, same contact base, with a persistent contact ID across waves. Transactional NPS is event-triggered — after a support ticket closes, an onboarding session ends, a feature releases. The follow-up question should change with the cadence; the rating question should not.
Relational NPS, no — anonymity loses the persistent contact ID and the trajectory across waves. Identified responses, attached to a persistent contact ID, are what makes the program useful for action. Transactional NPS can be anonymous if the team only needs to know about the event in aggregate, but most CS teams want to be able to route a detractor follow-up — which requires identification.
The same two. The eNPS rating question is "How likely are you to recommend [company] as a place to work?" on a 0-10 scale. The follow-up open-ended question varies but commonly asks "What is the primary reason for your score?" or "What would make this a 10?" The same survey-design discipline applies: keep the rating standard, design the follow-up carefully.
This page covers the survey instrument — what to ask, how to ask it, when to ask it. NPS analysis covers what to do with the answers. NPS feedback covers the closed-loop workflow around each response. NPS verbatim analysis covers the tool that reads the open-ended answers. The survey question is the start; the cluster covers everything after.
The survey is the start. The cluster covers everything after — the methodology, the verbatim analysis, the closed-loop workflow.
The pillar above this page — what NPS analysis means in the AI era, the methodology, the locked thesis behind the cluster.
What to do with the follow-up answers — the commercial decision about which tool reads the open-ended verbatim that this survey produces.
The full feedback signal that arrives with each response — the closed-loop workflow that turns the verbatim into action.
Your current follow-up wording, your last wave of NPS verbatims, your contacts. Sixty minutes. We read the verbatims your wording produced, walk through what a branched or anchored follow-up would have surfaced instead, and design the next quarter's question with the team. No demo accounts. No slideware. Your own survey, redesigned live.
No slideware. No demo accounts. Your own survey, redesigned live.
