NPS calculator

How the Net Promoter Score works

NPS comes from a single question that Fred Reichheld introduced in the Harvard Business Review in 2003: “How likely is it that you would recommend us to a friend or colleague?” People answer on a 0–10 scale and fall into three groups:

• Promoters (9–10): enthusiastic customers who are likely to recommend you.
• Passives (7–8): satisfied but not excited, and open to a competitor.
• Detractors (0–6): unhappy customers who may discourage others.

NPS = % promoters − % detractors. Take 200 responses: 120 promoters, 50 passives and 30 detractors. That is 60% promoters minus 15% detractors, an NPS of 45. Passives don't add or subtract, but they do count in the total, which is why moving a 7 to a 9 lifts the score.

Margin of error: how much your score could move

An NPS from a sample is an estimate. Scoring each response as +1 (promoter), 0 (passive) or −1 (detractor), the margin of error at 95% confidence is:

MoE = 1.96 × √( (P + D − (P − D)²) / n ) × 100

where P and D are the promoter and detractor shares as proportions and n is the number of responses. For the example above that is about ±10 points, so the true score is likely somewhere between 35 and 55. For a typical spread of 50% promoters, 30% passives and 20% detractors (NPS 30):

The practical lesson: a jump from 32 to 38 between two waves of 150 responses is well inside the noise. Don't celebrate it or panic about it. Compare the intervals, not the headline numbers.

What counts as a good NPS

A common rule of thumb reads below 0 as needs work, 0–30 as good, 30–70 as great and 70+ as excellent. Industry averages differ widely, and published benchmark reports disagree with each other because they survey different customers in different ways. The ranges in the calculator are indicative only. The most useful benchmark is your own score, measured the same way, over time.

The limits of NPS

• It throws information away.An 11-point scale collapses into three buckets, so a 6 and a 0 count the same, and moving someone from 0 to 6 doesn't change the score at all.
• Scale use varies by culture. Respondents in some countries rarely give 9s and 10s, so comparing NPS across regions can mislead.
• Who answers matters. Very happy and very unhappy customers respond more often. Survey a representative sample, at a consistent moment in the customer journey.
• It's a thermometer, not a diagnosis. Academic studies have questioned whether NPS predicts growth better than other satisfaction measures. The score tells you something is wrong; it never tells you what.

Relational vs transactional NPS

There are two common ways to run the survey, and their scores aren't comparable. Relational NPS asks your whole customer base at a regular interval, such as every quarter, and measures how people feel about the relationship overall. Transactional NPSasks right after a specific moment: a purchase, a support ticket, the end of onboarding. Transactional scores tend to swing with that single experience, so they're best for finding broken steps in the journey, while relational scores show the long-term trend. Pick one method per metric and keep the timing, channel and wording identical between waves, or you'll be measuring your process changes rather than your customers.

Making the score useful

Always ask a follow-up: “What's the main reason for your score?” The open-text answers from detractors are where the value is, and they usually point to a handful of fixable problems, often in onboarding, pricing clarity or a confusing part of the product. Close the loop with detractors personally, and track promoters and detractors separately as well as the net score.

For a ready-made survey and question wording, use my free NPS survey template. Testing a fix for what detractors told you? Check whether it worked with the A/B test calculator, and watch retentionalongside NPS. When the complaints point at the product itself, that's the work I do in product design.