The benchmark
SaaS NPS averages sit around 30-40 across most published industry surveys. A score of 50 or higher is excellent, and below 20 should be treated as a warning that something in the product or its audience fit is off. The score ranges from -100 to 100, calculated as the percentage of promoters (9-10 ratings) minus the percentage of detractors (0-6 ratings). Passives in the 7-8 range count against you by diluting promoters, which surprises most founders the first time they run the math.
What good looks like by segment
Products that end users choose for themselves, like developer tools and prosumer apps, often score 40-60 when they are working. Enterprise software imposed by procurement commonly sits at 10-30 even when renewal rates are strong, because the person answering the survey never chose the tool. Horizontal SMB tools tend to land near the 30-40 average. So a 25 can be solid for a compliance product sold to executives and mediocre for a tool developers picked themselves.
When this number lies
At 25 responses, each answer is worth 4 points of NPS, and the sampling error is roughly plus or minus 20 points, so a 45 and a 10 can reflect the same underlying sentiment. Response bias makes it worse: your happiest and angriest users answer surveys, and the silent middle, where churn lives, does not. Timing skews it too, since surveying right after onboarding or right after an outage produces different scores from the same product. Below a few hundred responses, read the free-text comments and ignore the number.
Retention beats NPS as a truth signal
People say 9 and then cancel; they say 6 and renew for years. Retention and expansion are what users do with their own money, which makes them the honest version of the question NPS asks. NPS is still useful as a cheap early-warning channel and a source of verbatim complaints, especially early on when you have little retention data. But when NPS and retention disagree, believe retention.