How to run an NPS survey at a startup

Net Promoter Score (NPS) is a loyalty metric built around one recommend question on a 0-10 scale, then a follow-up why. Bain & Company created the Net Promoter Score method with NICE Satmetrix; as of November 2026 Bain still documents the Net Promoter Score and Net Promoter System on its customer strategy pages. An NPS survey for startups should stay tiny: one score, one why, a send list you can staff, and fast survey replies when the score is low. The average is optional. Verbatim reasons and closed loops are what change the product.

NPS survey for startups starts with two questions

Ask only:

1. How likely are you to recommend us to a peer? (0-10)
2. What is the main reason for your score?

Add a third line only if every answer will be read: What is one thing we should improve next?

Skip feature wishlists, demographic mazes, and newsletter opt-ins on the same questionnaire. Length trains silence. Name the product people actually use, not a holding-company brand they barely recognize.

How the score is computed

Classify each 0-10 answer the standard Net Promoter Score way:

• 0-6: detractors
• 7-8: passives
• 9-10: promoters

NPS equals the percentage of promoters minus the percentage of detractors. Passives sit in the middle and are excluded from that subtraction. Print the sample size beside every NPS figure you share. With twenty replies, treat the figure as directional, not a board KPI.

Bain's Net Promoter materials emphasize loyalty and growth linkage for the broader Net Promoter System. Early teams should still treat a thin survey as a door into conversations, not as proof of product-market fit by itself. The method helps when the follow-up why and the closed-loop reply are funded; without those, the number becomes a vanity chart.

Wording that keeps answers honest

Prefer "recommend to a peer who has a similar job" over vague "recommend to a friend" when you sell B2B software. Peers share constraint language. Friends may score warmth instead of fitness for work. Keep scale labels visible: 0 = not at all likely, 10 = extremely likely. Do not invert the scale mid-year.

If you translate the survey, keep the 0-10 meaning intact. Do not collapse to a five-point smile scale and still call the result NPS.

Who receives the send

Invite people who completed a real workflow, not spectators who bounced after signup.

Write the invite in plain language. Example subject: "Two questions about (60 seconds)". State that low scores get a personal reply.

Do not buy a cold list and call it NPS. The method assumes a relationship worth recommending. Cold traffic measures brand curiosity, not loyalty.

Exclude employees and contractors from the external NPS pool unless you label an internal pulse separately. Mixing staff scores into external NPS hides real dissatisfaction.

Capture identity for closed loops

Collect work email and company if you do not already know the respondent. When they are known respondents, keep the person record in Contacts.

After each batch, triage into detractors, passives, and promoters for reply queues. With thin samples, do not treat the single NPS average as board-level truth.

File verbatims in Notes or a shared sheet with date, score, theme, owner, and reply status.

If you host the questionnaire in Nautis, Custom Forms includes an NPS (0-10) field type and an NPS survey template option. Keep this pulse as its own published form so exports stay readable. Builder steps for fields and live URLs belong in your forms playbook; this article stays on survey design and follow-through.

Reply to detractors within two business days

A low score without an survey reply extracts pain and then abandons the writer.

Reply note shape:

• Thanks for the score and the reason.
• Restate their reason in their words.
• Say what you will inspect and by when.
• Offer a short call only if it helps.

Promoters get a brief thanks. Separate any referral or case-study ask to a later, clear message. Passives get one focused question: what would have made this a nine?

Track closed-loop rate: detractors contacted within two business days divided by detractors received. If that ratio stays weak for two weeks, pause new sends until replies catch up.

Fund the reply queue before you enlarge the send list. Ten complete detractor replies beat a hundred ignored scores. Measure queue lag in hours, not vibes.

When a detractor names a concrete bug, attach the ticket or fix note in the same thread you use for closed-loop tracking. When they name pricing, do not promise a discount in the survey reply unless pricing already authorized it. Acknowledge the friction and route the commercial conversation to the right owner.

Invite copy that does not sound like marketing

Write the invite as an operations note. Name the product, the time cost, and what happens to low scores. Skip discount incentives; they change who answers. If completion stays under roughly thirty percent after two waves, fix the invite and timing before you add questions.

Monthly theme review

Group why-text into a short list: first-run friction, reliability, pricing clarity, support speed, missing workflow, other. Count mentions. Put a reviewer on the top two themes.

Honesty rules:

• Label segments when you compare averages (new vs long-tenure).
• Print sample size beside any number you share internally.
• Do not promote a roadmap change from a single angry line without checking frequency.
• Do not average unlike products or plans into one NPS without saying so.

When a theme repeats three times in a month, interview two of those respondents before you change packaging. The questionnaire finds the signal; the conversation confirms it.

YC's essential startup advice keeps pushing founders toward user conversations. Net Promoter Score is merely a thin, repeatable door into those talks, not a substitute for them.

What this survey is not

It is not a full discovery interview, not a win-loss program, and not proof of product-market fit by itself. An NPS survey for startups works when it yields dated verbatims, fast replies on low scores, and a monthly theme list with names attached.

Scoring pitfalls that waste a quarter

• Celebrating a rising NPS while closed-loop rate collapses.
• Comparing your score to a public consumer benchmark from another industry without context.
• Sending after every tiny in-app event until people mute you.
• Mixing employees, free-trial tire-kickers, and paying users in one unlabeled average.
• Deleting the why field to "raise completion" and then wondering why the number is useless.
• Treating passives as failures. Passives are a coaching pool, not automatic detractors.

Segment math without self-deception

If new activations dominate one quarter and long-tenure teams dominate the next, the NPS move may be mix, not improvement. Hold a simple cohort view: compute NPS for each segment, then optionally roll up. Publish both when leadership asks for "the number."

Keep a one-line definition of who counts as activated for your send. Without that line, each teammate invents a different sample and the quarterly chart stops matching reality.

Sampling bias to watch

People who love or hate you answer more often than quiet passives. That skews averages toward extremes. Weight your reading toward the why text and toward segments you care about commercially, not only toward whoever clicked first.

Week-one install

Stand up the two-question survey in whatever form tool you use (in Nautis, Custom Forms with an NPS field and a why field is enough). Send to ten qualified users. Reply to every detractor within two business days. Write a first theme list even if the sample is tiny.

When to halt a wave

Halt if answers collapse to one word, if people ask to stop mail, or if closed-loop rate is failing. Repair capacity before the next wave. Also halt if sales starts using the survey as a pressure channel for renewals; that contaminates honesty.

Relationship to other listening

Win-loss interviews explain why deals closed or died. First-run interviews explain early friction. This NPS pulse is a light, repeatable check across people who already reached real usage. Keep the instruments separate so insights stay attributable.

Internal reporting card

When you paste results for the team, lead with sample size, closed-loop rate, and top themes. Put the NPS average last. Teams that lead with the average argue about math; teams that lead with themes argue about work.

Share the raw why lines for the top theme in the same note. Summaries without quotes invite polite disagreement that never touches the respondent's words.

Promoter timing

Wait until a promoter has had a calm stretch before asking for introductions or a public quote. Keep the survey thank-you short; put any ask in a later message.

Cadence that survives a busy month

Quarterly for steady users plus a trigger after first meaningful workflow is enough for most early teams. Weekly blasts create survey fatigue and empty why fields. If you need faster signal, shorten the sample window on a fixed segment rather than spamming everyone.

Do not relabel other scores as NPS

Customer satisfaction (CSAT) and customer effort (CES) answer different questions. Mixing them into an "NPS dashboard" without the recommend question and the promoter or detractor math misstates the Bain method. If leadership wants a satisfaction pulse, run CSAT under its own name.

Because Bain and NICE Satmetrix defined promoters, passives, and detractors on the 0-10 recommend question, you should not invent alternate buckets and still label the chart NPS.