NPS = % promoters (9-10) − % detractors (0-6)
Passives (7-8) are counted in the base and then discarded from the result. That discard is where most of the confusion comes from.
The same score, opposite situations
Two products both score +20:
- A: 40% promoters, 20% detractors, 40% passive. Broad mild approval.
- B: 60% promoters, 40% detractors, 0% passive. Polarised.
These need completely different responses. A has an engagement problem; B has a segment that actively dislikes the product. The single number cannot tell them apart, so publish the distribution alongside it or the score is close to useless.
Sampling decides the number
Send a survey to everyone and the response rate will be 10 to 25%, skewed toward people with a strong opinion and toward the accounts your customer success team talks to most.
Two disciplines that matter more than the score itself:
- Sample consistently. Same trigger, same population, same timing. A score
compared across two different sampling methods is not a comparison.
- Report the response rate. An NPS of +45 on an 8% response rate from
hand-picked accounts is a testimonial exercise.
The follow-up is the metric
"Why did you give that score?" produces the only actionable output. The number tracks a trend; the verbatims say what to fix.
Code them into a small number of themes and report theme frequency beside the score. A team that reads verbatims monthly and ignores the score entirely is better off than one doing the reverse.
NPS and referral rate are different things
NPS measures stated willingness to recommend. Referral rate measures recommendations that arrived. Teams routinely treat the first as a proxy for the second, and the gap between them is usually large.
See referral rate.
A worked example
400 responses: 180 promoters, 140 passives, 80 detractors.
NPS = (180 ÷ 400) − (80 ÷ 400) = 45% − 20% = +25
The 140 passives, 35% of respondents, contribute nothing to the score while being the largest group and usually the most movable one.
The traps
- Publishing the score without the distribution or the response rate.
- Comparing across changed sampling methods.
- Comparing to a cross-industry benchmark, which is close to meaningless.
- Surveying only accounts your team has a relationship with.
- Treating it as a proxy for retention. It correlates, weakly.
Where this sits
NPS is Measurement in the Tenbound Pipeline Architecture Standard, and it is the softest number in the library. Read it beside a behavioural measure such as retention rate, which records what customers did rather than what they said.