What is NPS and is it actually useful?
Net Promoter Score asks how likely someone is to recommend you on a 0-10 scale, then subtracts the percentage of detractors (0-6) from promoters (9-10). It is useful as a tracked trend on a stable population and close to useless as a single number compared across companies, because the score moves with who you sampled and when.
What the number actually is
You ask one question — how likely are you to recommend this to a friend or colleague, 0 to 10 — and sort the answers into three buckets. Promoters are 9-10. Passives are 7-8. Detractors are 0-6. The score is the percentage of promoters minus the percentage of detractors, so it runs from -100 to +100.
Two things about that formula matter more than they look. The passives are discarded, and the detractor band is enormous: a 6 out of 10, which most people would read as mild approval, is counted the same as a 0.
Why cross-company comparison is mostly noise
An NPS of 42 means nothing without knowing who was asked and when. Survey people right after a successful onboarding and the score climbs. Survey the same population after a billing change and it collapses. Neither movement tells you the product got better or worse.
Cultural response patterns matter too — the same satisfaction level produces systematically different numbers in different markets, because willingness to give a 10 is not constant across the world. So a global product comparing its NPS to a published US benchmark is comparing two different things and calling the gap a finding.
The one way it earns its place
Track it on a fixed cadence, with a fixed sampling method, on your own product, and read only the trend. That version is genuinely useful: it catches slow deterioration that usage metrics hide, because people often keep using something they have started to resent.
And read the free-text box, not the score. The comment field attached to an NPS survey is usually worth more than the number it is attached to — it is a standing, low-effort channel for the specific complaints that would otherwise never reach you. Monzo's early growth ran heavily on word of mouth precisely because it treated that feedback loop as the product, not as a reporting obligation.
Where it actively misleads
The metric rewards being loved by a niche and punishes being useful to a majority. A product that solves a boring problem adequately for a large market will score worse than a polarising product with a devoted core, even when the first is the better business. WhatsApp spent a decade being unglamorous, universal and enormously valuable, which is not the profile NPS flatters.
Treat a low score as a question, not a verdict.
Seen in practice
Case studies where this shows up as a real decision, not a definition.
Related questions
What is a good NPS score?
There is no honest universal answer, which is the main problem with the metric. Published benchmarks vary by industry, by country, by survey timing and by how the question was asked, and the same product can score twenty points apart depending on whether you survey after a support interaction or at random. Compare your NPS to your own NPS last quarter, measured the same way.
Why does NPS throw away the 7s and 8s?
By design — the framework treats them as passive and excludes them from the calculation entirely. That is also why NPS is volatile on small samples: a handful of people moving between 8 and 9 swings the score even though almost nothing changed in what they think.
What should I use instead of NPS?
For product decisions, retention and repeat usage tell you more, because they measure what people did rather than what they said they might do. Superhuman's product-market-fit survey — how disappointed would you be if you could no longer use this — is a better early-stage signal because the answer is actionable.
More on metrics
Last reviewed 2026-09-08