Net Promoter Score (NPS) — How to Calculate, Implement, and Benchmark

Net Promoter Score (NPS) is the single most-used customer-experience metric in the world. The calculation: % Promoters minus % Detractors, based on responses to 'On a scale of 0–10, how likely are you to recommend us to a friend or colleague?' Scores 9–10 are Promoters; 7–8 are Passives; 0–6 are Detractors. NPS ranges from -100 to +100. Industry benchmarks vary widely — software typically 30+, airlines 0–20, telecom often negative.

NPS was introduced in 2003 by Fred Reichheld in HBR. The single-question survey captures customer advocacy likelihood and segments respondents into three groups with predictive value for repurchase, referral, and growth.

The exact calculation

  • Step 1: Send the NPS survey: 'On a scale of 0 to 10, how likely are you to recommend [Company/Product] to a friend or colleague?'
  • Step 2: Categorize each response: 0–6 = Detractor, 7–8 = Passive, 9–10 = Promoter
  • Step 3: Calculate percentages: %Promoters = (Promoters / Total Responses) × 100; %Detractors = (Detractors / Total Responses) × 100
  • Step 4: NPS = %Promoters − %Detractors (Passives are excluded)
  • Example: 200 responses; 100 Promoters (50%), 60 Passives (30%), 40 Detractors (20%). NPS = 50 − 20 = 30
  • Range: -100 (all Detractors) to +100 (all Promoters). Zero means equal promoters and detractors.

Industry benchmarks (median NPS by category)

  • Software/SaaS: 30–50 (top quartile: 50+)
  • Department/Specialty Stores: 50–65 (top retailers)
  • Tech/IT services: 30–40
  • Online retail: 40–55
  • Financial services: 20–40
  • Hotels: 30–55
  • Airlines: 0–20
  • Telecom/Cable: 0–10 (often negative)
  • Insurance: 20–35
  • Healthcare: 20–35
  • Top performers (any category): 70+

Statistical reliability and sample size

With 100 responses, NPS has roughly ±10-point margin of error at 95% confidence. With 1,000 responses, roughly ±3 points. Below 30 responses, NPS becomes statistically unreliable — directional only.

Margin of error formula: 1.96 × √[(σ²)/n] where σ is the standard deviation of NPS (typically 50–60 for diverse populations).

Transactional NPS (tNPS) vs Relationship NPS (rNPS)

  • Transactional NPS (tNPS): Sent immediately after specific event (purchase, support ticket, onboarding completion). Measures experience for that moment.
  • Relationship NPS (rNPS): Sent on regular cadence (quarterly, annual) capturing overall sentiment. Measures aggregate brand health.
  • Both serve different purposes — tNPS for operational improvement, rNPS for strategic measurement.

RGM Experts Say

NPS isn't the whole answer — but it is a useful signal when implemented well. The single most common mistake we audit: organizations celebrating NPS gains driven by changing the survey timing or audience, not real improvement. Always pair NPS score with response rate, audience, and timing. A score increase from 35 to 45 with declining response rate often means dissatisfied customers stopped responding, not that satisfaction improved.

Operating the NPS program

  • Survey timing: Avoid sending immediately after support interaction unless that's the question; bias toward over-reporting friction
  • Question wording: Use exact standardized wording for benchmarking
  • Follow-up question: 'What is the primary reason for your score?' — open-text
  • Detractor outreach: Within 24–48 hours, customer success outreach to detractors
  • Promoter activation: Invite to referral program, ask for review, request testimonial
  • Passive understanding: Survey passives for what would make them promoters
  • Closed-loop process: Track that detractor outreach happened and was logged
  • Trending analysis: NPS over time matters more than absolute number
  • Segment NPS: NPS by tier, plan, geography, tenure to find improvement levers

Related guides

Sources

  1. [1]Reichheld, The One Number You Need to Grow (HBR, 2003); Bain NPS benchmarks; Satmetrix NPS benchmarks