An NPS survey (Net Promoter Score survey) asks your customers one question about recommending you and turns the answers into a single number between minus 100 and plus 100. Setup takes a day, collecting takes about two weeks, and the follow-up comments tell you what makes people talk about you and what makes them warn others away. The score is a thermometer. The comments are the diagnosis, and this SOP spends most of its effort on them.

What is the NPS survey and when should you use it?

Fred Reichheld of Bain & Company introduced the Net Promoter Score in his Harvard Business Review article The One Number You Need to Grow. The method asks "How likely is it that you would recommend [company] to a friend or colleague?" on a 0 to 10 scale and sorts people into three groups:

Promoters (9 to 10): enthusiastic customers who will recommend you.
Passives (7 to 8): satisfied enough, but open to a competitor.
Detractors (0 to 6): unhappy customers who can hurt your reputation.

NPS = % promoters minus % detractors. Passives count toward the total but not toward either group. Net Promoter and NPS are registered trademarks of Bain & Company, NICE Systems and Fred Reichheld, and Bain's Net Promoter System site explains the calculation.

For early product-market fit work, the Sean Ellis test is the sharper tool, because it measures dependency. NPS earns its place once you have a steady customer base. Use this SOP:

When you have 100+ customers who used the product long enough to judge it.
When investors, partners or enterprise buyers ask for a score they recognise.
To track the effect of big changes (new pricing, a redesign, a support overhaul) across rounds.
To find the reasons people leave, before churn shows them to you.

What do you need before you start?

A customer list with segments: plan, signup month, company size or use case.
A minimum audience that can produce 100+ answers. At a 20% response rate, that means 500 customers.
A survey tool with a 0 to 10 scale and logic for follow-up questions.
A decision about type: relationship NPS (sent on a schedule about the whole relationship) or transactional NPS (sent after a specific event such as a support ticket). This SOP covers relationship NPS.

How do you run it, step by step?

1. Choose who to survey and when

Survey customers who have used the product for at least 30 days, so they judge the experience and not the first impression. Pick a fixed rhythm, every quarter or every six months, and avoid sending it right after an outage or a price rise unless you want to measure that.

2. Keep the form to two questions

Ask the standard recommendation question and one open follow-up: "What is the main reason for your score?" Every extra question lowers the response rate. Add a third only when you need it, such as "What is one thing we could do better?"

3. Send it and track the response rate

Email works best for B2B, in-app for products used daily. Send one reminder after four days. Record how many people received it, so the response rate is visible next to the score.

4. Calculate the score and its margin of error

Count promoters, passives and detractors, then subtract. Small samples swing wildly: with 50 answers, the 95% margin of error on NPS is roughly 20 points either way, and with 200 answers it shrinks to about 10. Never celebrate a 15-point jump measured on 40 answers.

5. Code the comments by group

Read every comment and tag it with a theme ("fast support", "missing export", "too expensive"). Count themes separately for promoters, passives and detractors. The promoter themes describe what people would repeat to a friend; the detractor themes describe what they would complain about.

6. Close the loop within a week

Reply personally to every detractor, thank promoters and ask them for a review or a referral. Net Promoter practitioners call this "closing the loop", and it is where most of the learning happens: a 15-minute call with a detractor teaches more than the score.

7. Write the round summary

One page: score, response rate, margin of error, top three themes per group, actions taken and the score by segment. Compare it with the previous round, never with a random industry average.

Template you can copy

NPS SURVEY (1 minute)

Email subject: One question about [Product]

Hi [first name],
We read every answer. Could you give us 30 seconds?

Q1. How likely is it that you would recommend [Product] to a friend
    or colleague?
    0  1  2  3  4  5  6  7  8  9  10
    (0 = not at all likely, 10 = extremely likely)

Q2. What is the main reason for your score? (open text)

Optional Q3. What is one thing we could do better? (open text)

SPREADSHEET COLUMNS
Customer ID | Segment | Months as customer | Score | Group (P/Pa/D) |
Reason | Theme | Follow-up done (Y/N) | Follow-up notes

SCORE CELLS
Promoters %  = COUNTIF(score, ">=9") / COUNT(score)
Detractors % = COUNTIF(score, "<=6") / COUNT(score)
NPS          = (Promoters % minus Detractors %) x 100

Worked example: a bookkeeping app for freelancers

This example is illustrative. A bookkeeping app has 900 paying freelancers. The founder surveys 700 who joined more than 30 days ago and receives 168 answers (24%).

Results: 71 promoters (42%), 52 passives (31%) and 45 detractors (27%). NPS = 42 minus 27 = +15. With 168 answers, the margin of error is about 12 points, so the true score sits between roughly +3 and +27.

The comments matter more. 38 of 71 promoters mention the automatic VAT report. 29 of 45 detractors mention bank sync breaking with two specific banks. The founder fixes the sync for those banks, emails every detractor who mentioned it, and asks promoters who praised the VAT report for a review. The next round, three months later, measures whether the sync detractors moved.

What should you measure, and when do you stop?

Metric	Pass signal	Warning signal
Net Promoter Score	Positive, and stable or rising across rounds	Negative, or falling two rounds in a row
Promoter share	40%+ of answers	Under 25%
Detractor share	Under 20%	Over 35%
Response rate	15%+ by email, 5%+ in-app	Under 5%: the score only reflects extremes
Reason themes	Half of promoters name the same reason	Promoter reasons are vague ("good app")

Stop a round when you reach your target number of answers or after ten days, whichever comes first. Stop using NPS as your fit signal if scores look fine while customers leave: compare it against customer retention analysis, because behaviour beats stated intent.

Which mistakes should you avoid?

Treating NPS as proof of word of mouth. Saying "I'd recommend you" costs nothing. Check real referrals with customer referral analysis.
Comparing with industry averages from blog posts. Sector benchmarks differ widely and samples are unknown. Compare with your own past rounds and direct competitors.
Reporting the score without the margin of error. Small samples produce noise that looks like progress.
Asking for a high score. "Anything below 9 counts as a fail for us!" corrupts the data.
Never calling detractors. The survey without the follow-up is half the method.

How can AI speed this up?

Export the comments and ask an AI assistant to tag each one with a theme from a fixed list you write, then count themes by group. Ask it to draft a personal reply for each detractor that references their exact complaint. Edit every reply before sending, because a templated apology to an angry customer reads as worse than silence, and spot-check 20 tags by hand to catch themes the AI merged.

For a comparison of survey and feedback tools that fit an early-stage budget, see the F/MS startup tool reviews.

Where does this SOP fit in your validation plan?

NPS sits in the Product-Market Fit stage, after the Sean Ellis test has shown that a segment depends on you. Related steps:

Check that promoters actually refer others with customer referral analysis.
Link detractor themes to churn with customer retention analysis.
Turn promoter themes into outcome tracking with customer success metrics.
If you need a broader survey later, follow the online survey validation SOP.

See all Product-Market Fit SOPs. The F/MS free startup school covers customer research and growth for first-time founders at no cost.

FAQ

What is a good NPS score for a startup?

Any positive score means more promoters than detractors. Track your trend across rounds and compare with direct competitors; published industry averages vary too much to serve as a target.

How many responses do I need for a reliable NPS?

At least 100. With 50 answers the score can be 20 points off in either direction; with 200 the range narrows to about 10.

What is the difference between NPS and the Sean Ellis test?

NPS asks whether customers would recommend you. The Sean Ellis test asks how disappointed they would be without you, which measures dependency and suits early fit decisions better.

How frequently should I send an NPS survey?

Every quarter or every six months for relationship NPS, and never to the same customer more than twice a year.

NPS Survey SOP