Where NPS comes from
Fred Reichheld introduced the metric in the Harvard Business Review article «The One Number You Need to Grow» in 2003, developed with Bain & Company and Satmetrix. The pitch was that one recommendation question predicts growth better than long satisfaction questionnaires. Whether it really predicts growth is still argued about, but the question stuck because it is short and comparable across teams.
Promoters, passives and detractors
| Score | Group | What it usually means |
|---|---|---|
| 9–10 | Promoters | Would recommend without being asked |
| 7–8 | Passives | Satisfied, but would switch for a better offer |
| 0–6 | Detractors | Unhappy enough to warn others |
Passives count towards the total but not towards either side, which is why moving people from 8 to 9 changes NPS more than moving them from 3 to 5.
NPS calculation formula with an example
NPS is written as a whole number without a percent sign. If you use Excel or Google Sheets, count answers with COUNTIF(range,">=9") and COUNTIF(range,"<=6"), divide each by COUNT(range) and subtract.
NPS = % promoters − % detractors
Example: 200 answers
Promoters (9–10): 90 → 45%
Passives (7–8): 60 → 30%
Detractors (0–6): 50 → 25%
NPS = 45 − 25 = +20What is a good NPS score?
Anything above zero means you have more promoters than detractors. Beyond that, benchmarks vary so much between industries, countries and survey channels that a universal «good» number does not exist. The honest comparison is with yourself: the same question, the same audience, the same moment in the customer journey, every quarter.
Always read the score together with the follow-up answers. A +30 with detractors complaining about one broken feature tells you what to fix; the bare number does not.
NPS survey question examples
- How likely are you to recommend [product] to a friend or colleague?
- How likely are you to recommend [company] as a place to work? (employee NPS)
- What is the main reason for your score? (open follow-up)
- What would we need to change for you to give us a 10?
When not to use NPS
- Right after a support ticket. You measure the ticket, not loyalty; ask a CSAT question instead.
- For products people do not discuss, such as internal tools with no alternative.
- With a handful of answers. Twenty responses can swing the score by 20 points either way.
Common NPS mistakes
- Changing the scale to 1–10 or 1–5. The groups only work on 0–10.
- Reporting the average score instead of NPS. An average of 7.8 hides whether you have fans or critics.
- Asking for a high score («we would love a 10»). It inflates the number and makes it useless.
- Skipping the «why» question. Without it you know the score and nothing else.
Running an NPS survey in Askyo
The NPS block is fixed to the 0–10 scale and comes with the question «How likely are you to recommend us?» and end labels «Never» and «Definitely». You only edit the wording and the two labels, so nobody on the team can accidentally break the scale.
On a phone the eleven buttons wrap into two rows (0–5 and 6–10) so each one stays easy to tap; on wider screens they sit in one line. The selected button turns red for 0–6, amber for 7–8 and green for 9–10.
The results page does the arithmetic: a donut with promoters, passives and detractors, the signed score in the middle and the 0–10 distribution. Put a long-text «why» question right after it and read the reasons next to the number.
What you can set
- label on the left end
- label on the right end