Net Promoter Score (NPS)

How the calculation works

Take every response to the 0–10 question and put it in one of three buckets.

With 200 responses — 90 promoters, 60 passives, 50 detractors — the maths is 45% promoters minus 25% detractors, which is +20. Not 20%, and not 20 out of 100. The passives are in the denominator both times, which is why a survey full of 7s and 8s can score zero.

What it is good at, and what it hides

NPS is good at one thing: giving a single comparable number that non-analysts will actually look at, tracked over time. Because it is a single item, it is cheap to ask and easy to append to any survey.

What it hides is most of the distribution. A move from +20 to +25 could be promoters gained, detractors converted to passives, or a change in who responded at all. The bucketing throws away the difference between a 0 and a 6, which are wildly different experiences, and treats 6 and 7 as opposites when they are one point apart. Always look at the raw distribution alongside the score, and always pair the number with an open-ended question asking why — the score tells you the temperature, the free text tells you the reason.

Common mistakes

• Reporting it as a percentage. It is an index on a −100 to +100 range. Writing "our NPS is 20%" is the fastest way to signal that nobody in the room checked.
• Calculating it on a tiny base. With 30 responses, one detractor moves the score by more than three points. Below about 100 responses the number is mostly noise.
• Changing the wording and calling it the same metric. "How likely are you to recommend us to a friend or colleague?" on 0–10 is the standard item. A five-point version, or a differently worded question, is a perfectly good measure — it is just not comparable to anyone else's NPS.
• Benchmarking against published industry figures. Different sampling, different populations, different points in the customer journey. Compare yourself to yourself.
• Ignoring who answered. If the survey goes out after a resolved support ticket, the sample is people whose problem got fixed. See sampling bias.

The name

"Net Promoter", "Net Promoter Score" and "NPS" are registered trade marks of Bain & Company, Fred Reichheld and NICE (formerly Satmetrix). The method is published and freely usable; the branding is not ours to put in a product interface. That is why NumoForms labels the question type Recommend score rather than NPS.

In NumoForms

The Recommend score question type collects a 0–10 answer with the standard eleven-point layout. Results show the distribution of answers per question, and the CSV export gives you one row per response, so bucketing into promoters, passives and detractors and taking the difference is a short formula in a spreadsheet. The product does not compute the index for you — and given how many teams want it segmented by branch, region or date, doing that arithmetic where your other data lives is usually the right place for it anyway.

For attitude measurement that keeps more of the distribution, see Likert scales. For the number that tells you whether the score is trustworthy at all, see response rate.