An NPS survey needs two questions: the Net Promoter Score question ("How likely are you to recommend us, 0 to 10?") and one open follow-up ("What is the main reason for your score?"). Adding more questions lowers response rates and works against the reason NPS exists, with a few deliberate exceptions covered below.
Key takeaways
- The standard NPS survey is two questions: the 0 to 10 score and an open "why".
- Fred Reichheld's team designed NPS specifically to replace long market research surveys that customers had stopped answering.
- The open answer replaces ten scored questions, because customers name what matters to them instead of rating what you guessed might matter.
- Extra questions are justified only with a clear knowledge strategy: you know what you will do with the answer.
How many questions should an NPS survey have?
An NPS survey should have two questions in almost every case: the recommendation score and the open reason. The debate is a hardy perennial in NPS practitioner communities, and it always follows the same arc: someone proposes adding "just three more questions" for another department, and a high-response survey quietly becomes a questionnaire nobody finishes. Protecting the survey from the organisation's wishlist is precisely the job. Every added question costs completions, and the completions you lose are biased: busy, mildly satisfied customers drop out first, leaving the extremes. The numbers are unambiguous: cross-platform completion research puts 4 to 8 question surveys at roughly 65% completion against about 42% for 15-plus questions, and SurveyMonkey's research shows dropouts multiplying once a survey passes the 7 to 8 minute mark. The trade-offs behind survey length are a topic of their own in our piece on response rates.
What is the exact NPS question?
The NPS question is: "On a scale from 0 to 10, how likely are you to recommend [company] to friends, family or colleagues?" For B2B, "colleagues or business partners" replaces "friends or family". Keep the scale visible as 0 to 10 with labelled endpoints, and never reword "recommend" into "rate" or "score"; the recommendation frame is what makes results comparable and is the basis of the whole Net Promoter system.
Which follow-up question should you ask?
The follow-up is one open question: "What is the most important reason for your score?" Open, singular and neutral. Not "what did you like?", which fishes for praise, and not a checklist of predefined topics, which limits customers to your assumptions. The open format is what makes two questions sufficient: the verbatim carries the diagnosis.
Which follow-up variants work for specific segments?
Within the two-question discipline, the open question can be tuned by segment once the answer arrives, in the follow-up conversation rather than the survey itself. For detractors, the reply that works asks one thing: "What would we need to change for you to score us higher?", which converts a complaint into a specification. For promoters, the follow-up is the referral or review moment: a customer who just scored you 9 is at peak willingness to say so publicly, and routing them to a review platform turns private loyalty into visible proof. For passives, the useful probe is "What is the one thing that kept this from being a 9?", because passives usually have exactly one reason and it is usually fixable. Note the pattern: none of these lengthen the survey. They live in the close-the-loop workflow after the response, where a human (or a well-configured automation) asks the next question only of people who already answered the first two.
Why only two questions? The four arguments
- Response rate is the ceiling on everything. A feedback programme's statistical value depends on volume and representativeness, and both fall with every extra field. Short surveys are also answerable on a phone in a queue, which is where your customers are.
- The philosophy: NPS was launched to fight exactly the long surveys and market research tools that customers had grown tired of. Reichheld's team built one magical question because people had stopped filling out the twenty. Re-inflating it with extra questions rebuilds the problem it solved.
- The open answer outperforms your question list. Scored questions can only confirm or deny what you thought to ask. The verbatim tells you what you didn't think to ask, in the customer's own priority order.
- Analysis is no longer the bottleneck. The historical reason for many scored questions was that open text was unanalysable at scale. AI text analysis such as ISAAC classifies every open answer by topic and sentiment and key driver analysis links themes to score movement, so two questions now yield more insight than fifteen did.
When are more questions justified?
More NPS survey questions are justified when a knowledge strategy demands them: you know exactly what decision the extra answer feeds. Legitimate cases: one demographic question when your CRM genuinely lacks the segment, one targeted driver question during a strategic change ("How do you rate our new delivery option?"), or a permission question for follow-up contact. The discipline is one extra question at most, with a defined owner for the answer. What never belongs in an NPS survey: satisfaction batteries duplicating what CSAT and CES already measure at their own touchpoints.
What do bad NPS surveys look like?
Bad NPS surveys share recognisable anti-patterns, and each one damages the data in a specific way. Matrix batteries ("rate the following twelve aspects") turn a one-minute survey into a chore and produce straight-lined answers where every row gets the same score. Mandatory comment fields force filler text ("all fine") that pollutes text analysis. Double-barreled questions ("how satisfied are you with our speed and service?") produce answers you cannot attribute. Leading introductions ("we work hard every day for you; how likely...") inflate scores and destroy trend value. Demographic questions your CRM already answers waste the customer's goodwill on data you own. And over-surveying, the quietest killer, trains customers that your emails are safe to ignore. The test for every proposed addition is one sentence: name the decision this answer will change and the person who will make it. No answer, no question.
How should you deliver the NPS survey: channel, timing and sampling?
Delivery decisions move NPS results as much as wording does. Channel: email remains the workhorse for relationship NPS; in-app surveys respond faster but skew toward active users; SMS earns its higher cost in low-email industries. Whichever you choose, keep it constant, because switching channels mid-programme moves scores by itself. Timing: send relationship surveys on a rolling schedule spread across the base rather than in one quarterly blast, so seasonality and campaign noise average out. A reminder two to three days after the initial invitation lifts response meaningfully (Clootrack's benchmark review puts typical survey response at 20 to 30%, with reminders adding up to a third more answers). Sampling: quarantine every respondent for at least a quarter, and make sure the sample mirrors your customer base rather than your most reachable segment, or the score describes your email list instead of your business. Route every detractor answer into a close-the-loop workflow the moment it arrives; the survey and the follow-up are one system.
NPS survey composition by use
| Survey type | Questions | Notes |
|---|---|---|
| Relationship NPS | Score + open why | The default; quarterly or continuous rolling |
| Transactional follow-up | CSAT or CES + open why | Touchpoint metrics, not NPS |
| Strategic deep-dive | Score + open why + 1 targeted question | Time-boxed, with a knowledge strategy |
| B2B account NPS | Score + open why (+ contact permission) | Route detractors to close-the-loop |
FAQ about NPS survey questions
What are the two NPS survey questions?
The score question, "On a scale from 0 to 10, how likely are you to recommend us to friends, family or colleagues?", and the open follow-up, "What is the most important reason for your score?"
Can you add extra questions to an NPS survey?
Yes, but sparingly and with a defined purpose. One extra targeted question with a clear owner is defensible; a battery of satisfaction items is not, because response rates and data quality fall with length.
Why does NPS use a 0 to 10 scale?
The 11-point scale allows the standard promoter (9-10), passive (7-8) and detractor (0-6) segmentation and keeps scores comparable across companies and industries. Changing the scale breaks comparability.
Should the open question be mandatory?
No. A mandatory text field lowers completion and produces filler answers. Voluntary open answers are fewer but far more diagnostic.
How do you analyse the open answers at scale?
With AI text analysis that classifies each answer by topic and sentiment, then links topics to score impact. Manual reading works below a few hundred responses a month; beyond that it becomes inconsistent.
Does the order of the two questions matter?
Yes: score first, open question second. The score primes a specific memory of the experience, which makes the open answer sharper. Reversing the order lowers completion because a text box as the opening ask feels like work.
---