45 NPS Questions (and the Follow-Ups an AI Would Ask)

The NPS question itself is the easy part — there's exactly one, and its wording barely changes:

"How likely are you to recommend us to a friend or colleague?" — answered on a 0–10 scale.

Everything that makes an NPS programme useful happens after that question. The score sorts your customers into detractors (0–6), passives (7–8), and promoters (9–10) — but the score alone tells you what happened, never why. The value is in the follow-up, and the follow-up is where most NPS surveys fall flat: a single optional comment box that most people skip.

So here's the question bank: the core question and its variants, then the follow-ups worth asking each score band — and, at the end, the honest problem with scripting follow-ups at all.

The core NPS question (and when to change it)

1. "How likely are you to recommend [company] to a friend or colleague?" — the standard wording, from Fred Reichheld's original 2003 Harvard Business Review research ("The One Number You Need to Grow"). Use it unmodified if you want comparability with published benchmarks.
2. "How likely are you to recommend [product] to a friend or colleague?" — product-level NPS, for companies with more than one product.
3. "How likely are you to recommend [company] to someone in a similar role?" — a B2B variant; "friend or colleague" can feel off for procurement-heavy purchases.
4. "Based on today's experience, how likely are you to recommend us?" — the transactional variant, tied to a specific interaction (a purchase, a support ticket, an onboarding call) rather than the whole relationship.
5. "How likely are you to recommend working with our team?" — the services variant, for agencies and consultancies where the team is the product.

Keep the 0–10 scale in every case. Change the wording and you can still learn plenty; change the scale and you can't compare with anything, including your own history.

Follow-up questions for detractors (0–6)

A detractor score is an opened door — the worst thing you can do is respond with silence. Ask:

1. "What's the main reason for your score?"
2. "What happened? Walk us through the last time we let you down."
3. "What were you trying to get done when things went wrong?"
4. "What's the one thing we could change that would raise your score?"
5. "Was this a one-off problem, or has it happened repeatedly?"
6. "What do you use alongside (or instead of) us to get around the problem?"
7. "Did you contact support about this? How did that go?"
8. "What almost made you choose a different provider — or what's keeping you from switching now?"
9. "If you were in charge here for a week, what would you fix first?"
10. "How has this affected your day-to-day work (or your team's)?"
11. "Is there anything we could do to win back your confidence in the next 90 days?"
12. "Would you be open to a follow-up conversation once we've addressed this?"

Follow-up questions for passives (7–8)

Passives are the most information-rich band and the least questioned. They're satisfied — and satisfied is exactly what a competitor can beat. Ask:

1. "What would it take to make that a 9 or a 10?"
2. "What's working well — and what's merely fine?"
3. "What almost frustrated you recently, even if it worked out?"
4. "Which part of the product do you get the most value from? The least?"
5. "Is there a job you wish we handled that you currently do elsewhere?"
6. "What would you miss most if you stopped using us tomorrow?"
7. "Have you recommended us to anyone? What stopped you, or what did you say?"
8. "If our price rose 20% tomorrow, what would you do?"

Follow-up questions for promoters (9–10)

Promoters tell you what to protect, what to amplify, and — asked well — hand you the words for your marketing. Ask:

1. "What's the main reason for your score?"
2. "What do you tell people when you recommend us?"
3. "What problem were you trying to solve when you found us — and how was it going before?"
4. "What nearly stopped you from signing up?"
5. "Which feature or moment turned you into a fan?"
6. "Who else do you think should be using this, and why?"
7. "What do we do better than anything else you've tried?"
8. "Is there anything that, if we broke it, would cost us your 10?"
9. "What should we build next for someone like you?"
10. "Would you be happy for us to quote you — or introduce you to a customer weighing us up?"

Contextual NPS questions for specific moments

1. "Now that you've finished onboarding, how likely are you to recommend us?" — catches first-impression problems while they're fresh.
2. "Following your support conversation, how likely are you to recommend us?" — transactional NPS for the support team.
3. "As you approach renewal, how likely are you to recommend us?" — the score that predicts the meeting.
4. "After [major release], how likely are you to recommend us?" — did the big bet land?
5. "How likely is your team to recommend us?" — for B2B accounts where the buyer and the users are different people.
6. "How likely are you to recommend us for [specific use case]?" — when your product serves clearly distinct jobs.
7. "Compared with a year ago, are you more or less likely to recommend us — and what changed?" — trend in a single question.
8. "You scored us [X] last quarter and [Y] now. What changed?" — the follow-up almost nobody asks, because forms can't remember.
9. "Before you go: how likely would you have been to recommend us at your happiest — and when was that?" — the exit variant, for offboarding flows.
10. "Is there anything you'd like to tell us that we didn't ask about?" — always last, always open.

The honest problem: you can't script your way to "why"

Look back at that list. The best questions on it — "what changed?", "walk us through what happened", "what almost frustrated you?" — aren't really questions. They're the first move in a conversation. The real information arrives in the answer to the answer: someone says "support was slow", and everything useful lives one level down. Slow how? Once or always? What did it cost you?

A form can't go one level down. Whatever you script, the respondent gets the same next question regardless of what they just said — which is why NPS comment boxes fill up with "it's fine" and "just make it better", and why analysing them becomes an afternoon of tagging spreadsheet cells.

This is the case for making the follow-up conversational. In Flowback, an NPS campaign starts with the same 0–10 question — but what follows is an AI conversation, not a comment box. You set the goal ("understand what's driving detractor scores this quarter") and anchor questions; the AI asks the natural follow-up based on what each person actually says, promoter or detractor, and keeps going until the reason is on the table. Then the part you'd normally do by hand happens automatically: every response is summarised, the score aggregates sit next to an AI summary of the verbatims, recurring reasons cluster into themes across campaigns — and the ones that need action become tracked issues, with two-way Linear, Jira, and GitHub sync, so "support was slow" ends as a fix someone shipped, not a row in a spreadsheet.

The questions above are a solid script. The better move is to stop scripting: pick the core question, set the goal, and let the conversation find the why. You can start an NPS campaign in Flowback free for 14 days.

Frequently asked questions

What is the NPS question?

"How likely are you to recommend us to a friend or colleague?", answered on a 0–10 scale. Respondents scoring 9–10 are promoters, 7–8 are passives, and 0–6 are detractors. Your Net Promoter Score is the percentage of promoters minus the percentage of detractors, giving a score from −100 to +100.

Can I change the wording of the NPS question?

You can adapt the wording to your context — product-level, transactional, B2B — but keep the 0–10 scale and the "how likely to recommend" framing. The further you drift from the standard wording, the less comparable your score is with benchmarks and with your own history.

Should NPS surveys include follow-up questions?

Yes — the score without the reason is close to useless for deciding what to do next. At minimum, ask "what's the main reason for your score?" Better, tailor follow-ups by band: detractors get diagnosis questions, passives get "what would make it a 10?", promoters get "what do you tell people?" Best, make the follow-up a conversation that adapts to each answer.

How often should you run NPS?

Relational NPS (the overall relationship) is typically run quarterly or twice a year; transactional NPS fires after specific moments like onboarding or a support interaction. The common failure is over-surveying the same people — space it out and always show that previous feedback led somewhere.

Is NPS still worth running?

The score is a blunt instrument, and treating it as a KPI to game helps nobody. But as a conversation starter — a low-friction way to sort customers into "something's wrong", "unconverted", and "something's working", and then ask why — it has no simpler rival. The value is almost entirely in the follow-up.

Net Promoter®, NPS®, and the NPS-related emoticons are registered trademarks of Bain & Company, Inc., NICE Systems, Inc., and Fred Reichheld. The core question originates in Reichheld's "The One Number You Need to Grow", Harvard Business Review, December 2003.