Respondents scoring 7-8 are 'passives' and are left out of the calculation entirely. They're neither subtracted nor added, which surprises people seeing the formula for the first time.
NPS is a single number with no context of its own. A score of 40 means nothing without knowing your industry's typical range, your own trend over time and, most importantly, why detractors are detractors, which the score itself never tells you.
The number is only useful paired with a follow-up 'why' question. Without it, NPS is a scoreboard with no explanation for the score, and teams end up debating the number instead of acting on the cause.
Formula
NPS = %Promoters (score 9-10) − %Detractors (score 0-6)
Passives (score 7-8) are counted in the response total but don't appear on either side of the subtraction. It's a common source of manual calculation errors.
Why NPS matters
NPS correlates with organic referral behavior and repeat-purchase likelihood, making it one of the few survey metrics that plausibly connects to actual revenue, not just sentiment. Tracked over time, a moving NPS is often an earlier warning sign of churn risk than churn-rate itself, since it's forward-looking, not a lagging count of who already left.
Same average score, different NPS
100 respondents split 40 promoters, 40 passives, 20 detractors gives an NPS of 20 (40% − 20%). A different 100 respondents split 60 promoters, 0 passives, 40 detractors also averages out similarly on a mean score, but gives an NPS of 20 (60% − 40%) from a far more polarized base. Two very different customer bases producing the same headline number, which is exactly why the raw score without distribution is easy to misread.
Common mistakes
Treating NPS as comparable across industries without adjustment
'Good' NPS varies hugely by category. Subscription software and airlines have very different typical ranges. Benchmarking your score against an unrelated industry's average produces a meaningless comparison.
Skipping the follow-up 'why' question
A score with no reason attached gives you a temperature reading with no diagnosis. The follow-up question is where the useful information actually lives. The number alone just tells you something is or isn't working.
Surveying only happy, engaged customers
Sending NPS surveys only to customers who recently had a good support interaction, or excluding churned customers from the sample, inflates the score and hides the detractor feedback that would have been most useful.
Where we work on this