Net Promoter Score gets treated in a lot of SaaS dashboards as a headline metric next to MRR and churn, and I think that's the wrong altitude for it. NPS is a survey — a single question, a self-reported number, collected at whatever moment you happened to ask — and it behaves like one: noisy, laggy, and easy to game by asking the wrong people at the wrong time. It's still worth collecting. It's just not worth treating as ground truth on its own.
How it's calculated
You ask "How likely are you to recommend [product] to a friend or colleague?" on a 0-10 scale. Respondents scoring 9-10 are Promoters, 7-8 are Passives, 0-6 are Detractors. NPS = %Promoters − %Detractors, giving you a number from -100 to 100. Passives don't count in either direction — they're excluded from the formula entirely, which surprises people the first time they compute it by hand.
SELECT
ROUND(
100.0 * SUM(CASE WHEN score >= 9 THEN 1 ELSE 0 END)
/ COUNT(*)
-
100.0 * SUM(CASE WHEN score <= 6 THEN 1 ELSE 0 END)
/ COUNT(*)
, 1) AS nps
FROM survey_responses
WHERE survey_date >= CURRENT_DATE - INTERVAL '90 days';Survey timing is the part people get wrong
When you ask matters as much as who you ask. Survey someone the day after onboarding and you're measuring first-impression enthusiasm, not product satisfaction. Survey only your most engaged power users — the ones who log in daily and open support tickets because they care — and you'll systematically miss the silent, dissatisfied users quietly planning to churn at renewal. A common pattern is triggering the survey at a fixed point in the customer lifecycle (30 days post-onboarding, then quarterly) rather than in-app randomly, so responses are comparable across cohorts instead of skewed toward whoever happened to be active that week.
If only 8% of surveyed customers respond, and the respondents skew toward people with strong opinions in either direction, your NPS is measuring "opinions of engaged extremes," not "sentiment of your customer base." A low response rate should make you trust the number less, not more — and it's worth reporting response rate alongside NPS, not just the score.
Why it's a lagging, blunt signal
NPS tells you how someone felt when they answered the survey, which might be weeks or months before they actually churn or expand. It's also blunt: a 7 and a 3 are both "Detractors" in the binary sense that matters for the score, but they're very different customers — one is mildly annoyed, the other is actively looking at competitors. The single number collapses that distinction. If you're using NPS as an early-warning system for churn, you're better off looking at the distribution of scores and, more importantly, the free-text comments that usually accompany the score — the comment is where the actionable signal actually lives.
Correlating it with churn and expansion
NPS becomes more useful when you stop treating it as a standalone metric and start joining it against what customers actually do. Pull the NPS score for each account, then check: did Detractors churn or downgrade at a materially higher rate over the following two quarters than Promoters? If the correlation is weak or absent in your own data, that's a sign your survey timing, sample, or question wording isn't capturing anything predictive for your business — not every SaaS company's NPS correlates cleanly with retention, and it's worth checking rather than assuming.
A blended NPS across self-serve and enterprise accounts hides more than it reveals — enterprise customers with a dedicated CSM tend to score differently than self-serve users who've never talked to a human. Break NPS out by customer segment before drawing conclusions about "how customers feel" as a single population.
Common mistakes worth naming
The recurring ones: surveying at a moment that flatters the product (right after a big feature launch or a successful onboarding call) instead of a neutral, consistent point in the lifecycle; treating month-to-month NPS swings as meaningful when the sample size per period is small enough that the swing is just noise; and reporting NPS to a board or investor as if it were a leading indicator with the same rigor as MRR, when it's a self-reported survey with a single-digit response rate. None of these make NPS worthless — they just mean it needs the same skepticism you'd apply to any small-sample survey data.
Wrapping up
NPS is useful as one input among several, especially when you look at the trend over time and the comments behind the score rather than the number in isolation. It's not a substitute for retention data, and a single quarter's NPS move shouldn't change a product roadmap on its own — treat it as a smoke detector, not a thermometer.
Independent software engineer in Nairobi specialising in Acumatica customisations, Laravel backends, and tax fiscalisation integrations across East and Southern Africa.