NPS calculator

Net Promoter Score is a loyalty metric built on a single question: "How likely are you to recommend us to a friend or colleague?", answered on a 0-10 scale. It was introduced by Fred Reichheld and Bain & Company in 2003, and it has become the most widely used customer loyalty measure in business.

Its appeal is that it compresses a messy question — how do people feel about us? — into one number between −100 and +100 that anyone in the organisation can follow over time.

The three groups

• Promoters (9-10) — enthusiastic, loyal, and the source of most referrals. They stay longer and cost less to serve.
• Passives (7-8) — satisfied but indifferent. They won't complain, but they'll switch for a better offer without much hesitation.
• Detractors (0-6) — unhappy. They churn, and they tell other people why.

The cut-offs look arbitrary but aren't: they came from Reichheld's research into which score ranges actually predicted repeat purchases and referrals.

Convert each group to a percentage of your total responses, then subtract detractors from promoters:

Step by step

1. Count responses in each band: 9-10, 7-8, and 0-6.
2. Add them together to get your total response count.
3. Divide promoters by the total and multiply by 100.
4. Do the same for detractors.
5. Subtract the detractor percentage from the promoter percentage.

Worked example

You collect 200 responses: 120 promoters, 40 passives, and 40 detractors.

Your NPS is +40. Note the result is a plain number, not a percentage — writing "40%" is technically wrong, though extremely common.

Bain & Company, who created the metric, publish these broad interpretation bands:

Why the absolute number matters less than the trend

NPS is strongly affected by context that has nothing to do with your performance: region, industry norms, how and when you ask, and whether the survey follows a support interaction or arrives out of the blue. Comparing your score against a competitor's published figure usually compares methodology, not loyalty.

What NPS does well is track change against a consistent method. Keep the question, the scale, the timing, and the audience stable, and movement in the score becomes a signal worth acting on.

There are two ways to run the survey, and they answer different questions.

Relational NPS

Sent on a fixed cycle — quarterly or twice a year — to a broad sample of customers, regardless of recent activity. It measures overall sentiment toward the brand and is the right choice for a board-level trend line.

Transactional NPS

Triggered by a specific event: a support ticket closing, an order arriving, an onboarding finishing. It measures that experience while it's fresh, and it's far more actionable because you know exactly what the respondent is reacting to.

Most mature programmes run both, and keep the two scores separate. Blending them produces a number that moves for reasons you can't diagnose.

• Always ask the follow-up. "What's the main reason for your score?" as an open text field turns a number into something you can act on. It is the single most valuable question in the survey.
• Keep the wording fixed. Changing the question, the scale, or the labels breaks comparability with every previous wave.
• Watch your sample. If only your most engaged customers respond, your score is inflated. Check that respondents resemble your customer base — and use the sample size calculator to work out how many responses you need first.
• Don't survey the same people too often. Quarterly is plenty for relational NPS. Monthly produces fatigue and falling response rates.
• Close the loop. Follow up with detractors. Programmes that do this consistently see scores improve; programmes that only measure don't.
• Segment before you conclude. A flat overall score can hide a plan tier or region that's falling apart.

NPS is useful, but it's worth knowing what it can't do. It's a single number, so very different distributions can produce identical scores — 50% promoters and 10% detractors gives +40, and so does 45% promoters with 5% detractors, despite meaningfully different customer bases.

It also measures stated intent rather than behaviour, and stated intent to recommend is an imperfect predictor of actual recommendation. And because it collapses to one figure, it tells you nothing about why without a follow-up question.

Use it as one input alongside CSAT for specific interactions, plus retention and churn data for what customers actually did.