Net Promoter Score (NPS)
One question, one number, wide adoption
Net Promoter Score reduces customer sentiment to a single question — "On a scale of 0 to 10, how likely are you to recommend [company] to a friend or colleague?" — and a single resulting number. Its appeal is exactly that simplicity: it's cheap to run, easy to trend over time, and comparable across companies and industries in a way most satisfaction metrics aren't.
That same simplicity is also its biggest weakness, and understanding both halves is what separates teams that use NPS well from teams that over-trust it.
The survey methodology
A standard NPS survey has two parts: the 0–10 likelihood-to-recommend question, followed immediately by an open-text follow-up — "what's the primary reason for your score?" The second question is not optional if the score is going to be useful; the number tells you *that* sentiment moved, the open text tells you *why*, and without it NPS becomes a number nobody can act on.
Timing matters as much as wording. Relationship NPS is sent periodically (quarterly or biannually) to a broad customer base and measures overall sentiment toward the brand. Transactional NPS is triggered immediately after a specific interaction (a support ticket, an onboarding flow, a purchase) and measures sentiment toward that moment specifically. Conflating the two — treating a transactional dip as a relationship-wide problem, or vice versa — leads to the wrong team fixing the wrong thing.
The promoter / passive / detractor math
Respondents are sorted into three buckets based on their 0–10 score, and the buckets are deliberately uneven in size:
- Detractors (0–6) — unhappy enough to actively discourage others, or churn-risk at minimum.
- Passives (7–8) — satisfied but unenthusiastic; they won't actively promote you, and will switch for a marginally better alternative.
- Promoters (9–10) — enthusiastic enough to actively recommend you.
The score itself is: NPS = %Promoters − %Detractors. Passives are counted in the denominator but don't appear in the subtraction, which is the detail most people get wrong when they try to recompute the score from raw averages.
Survey responses (n = 200):
Scores 0-6 (Detractors): 40 respondents → 20%
Scores 7-8 (Passives): 60 respondents → 30%
Scores 9-10 (Promoters): 100 respondents → 50%
NPS = %Promoters - %Detractors
NPS = 50% - 20% = 30
Range is always -100 (all detractors) to +100 (all promoters).
Above 0 is generally considered acceptable; 50+ is strong;
70+ is rare and typically limited to category-leading brands.NPS score bands and rough interpretation
| Score range | General interpretation | Caveat |
|---|---|---|
| Below 0 | More detractors than promoters | Investigate immediately, but check industry baseline first |
| 0 – 30 | Acceptable, room to improve | Typical for many B2B and utility categories |
| 30 – 50 | Good, healthy promoter base | Still worth segmenting by cohort for decay |
| 50+ | Excellent, rare outside leading consumer brands | Verify survey isn't biased toward happy respondents |
There is no universal good score — only a category baseline
NPS's limitations as a single metric
The core criticism of NPS, well-documented since its popularization, is that a single question collapses too much nuance into one number to drive specific action. A dropping NPS score tells you sentiment is worsening; it does not tell you whether that's a pricing problem, a support problem, a product-quality problem, or a one-time cohort of unhappy customers skewing the sample. Teams that treat NPS as the whole measurement system rather than one input tend to chase the number directly — incentivizing survey timing, excluding unhappy segments from the send list, or over-indexing on follow-up text from the loudest respondents — instead of fixing the underlying experience.
It's also sensitive to sampling and timing in ways that are easy to miss: response rates for NPS surveys are typically low (single digits to low teens), and non-response is rarely random — people with strongly negative or strongly positive experiences are more likely to respond than people in the middle, which can push the measured score in either direction relative to the true population sentiment.
NPS vs. complementary satisfaction metrics
| Metric | What it captures | Best paired with |
|---|---|---|
| NPS | Long-term loyalty / likelihood to advocate | Relationship-level health tracking over time |
| CSAT | Satisfaction with a specific interaction | Transactional touchpoints (support ticket, checkout) |
| CES (Customer Effort Score) | How much effort a task required | Diagnosing friction in a specific flow |
| Customer Health Score | Composite predictive risk signal | Proactive churn/expansion intervention, not just sentiment |
The trend matters more than the absolute number
What's next
NPS is one input into a broader read of customer risk; combining it with usage and support signals into a single predictive score is the next step up in sophistication.
Next: Customer Health Scoring →
I build these systems professionally.
Whether it's a RAG pipeline, analytics migration, or AI workflow — let's talk.