6.2 Net Promoter Score (NPS) Survey

In Brief

Description

A Net Promoter Score (NPS) survey is a single-question loyalty survey that asks people who have used your product one fixed question — how likely they are to recommend it — on a 0-to-10 scale, then converts the answers into one score you can track over time. It is meant to measure loyalty and identify the customers most likely to recommend you to others.

The question wording does not change: ask “How likely is it that you would recommend [company / product] to a friend or colleague?” on a 0-to-10 scale. Sort the answers into three groups — Promoters (9-10, your most enthusiastic customers), Passives (7-8, satisfied but unexcited), and Detractors (0-6, unhappy). The score is the percentage of Promoters minus the percentage of Detractors, which lands somewhere between -100 and +100. Keeping the wording and the brackets fixed is what lets you compare one round against the next.

The number itself is only an anchor; the value comes from the open-ended “why” question you ask alongside it. Those written answers are where the useful themes live — what promoters praise, and what detractors want fixed.

The score tells you about past satisfaction, not future behavior, so pair it with what customers actually do — how much they use the product, who is at risk of leaving, who refers others — rather than reading it in isolation. Teams also tend to game NPS when they tie bonuses to it, chasing the number instead of acting on the feedback; treat the score as a prompt to do the follow-up work, not a target to hit.

How to

Prep AI Prompt

1. Write down why you’re sending this NPS. “Track quarterly trend,” “find detractors before they churn,” and “benchmark against competitors” require different cadences and follow-up workflows. If you can’t name the decision the score will inform, you’re not ready to field it.

2. Decide who to survey. Pull the list from your CRM or product analytics, filtering for users past onboarding and active recently enough to have a real opinion. Gather enough qualified responses for the score to be meaningful — sampling noise stays large until you clear roughly 30 replies (see Analysis) — and if your user base is small, survey everyone who qualifies.

3. Decide cadence. Pick one and commit. Relationship NPS surveys the same cohort every 90 or 180 days and tracks the trend of overall loyalty. Transactional NPS is triggered by a specific event (purchase, support ticket close, feature first-use) and diagnoses which moments produce promoters and which produce detractors. Running both at once without segmentation muddies the signal.

4. Write the survey. The core question is standardized — don’t change it:

   • “How likely are you to recommend [product] to a friend or colleague?” (0 = Not at all likely, 10 = Extremely likely)
   • Add one open-ended follow-up: “What is the primary reason for your score?”
   • Optionally: “What could we do to improve your experience?”
   • Keep the NPS question first. If you embed NPS in a longer survey, the score is not comparable to standalone NPS.

5. Plan the close-the-loop workflow before you send. Decide who will contact every detractor and within what window (Bain’s standard is 24–48 hours). If nobody is on the hook to respond, don’t run the survey — collecting feedback you won’t act on burns goodwill with customers most likely to be saved.

Analysis AI Prompt

The original NPS calculation, per Reichheld and Bain, subtracts the percentage of Detractors from the percentage of Promoters and ignores Passives:

NPS = % of Promoters – % of Detractors

Before you read too much into the number, remember it is an estimate from a sample, not a census, so it carries a margin of error that shrinks as responses grow. At around 30 responses, sampling noise alone can swing the score by 10 points or more, so treat early readings as directional. A wave-over-wave change is only statistically meaningful when it is larger than that margin — with the small samples typical of early-stage startups, that usually means a move of several points on a few hundred responses, not a one- or two-point wobble on thirty.

A single score doesn’t mean much on its own. Anchor it three ways:

• Against your prior wave. A 5-point movement across consecutive quarterly waves with the same cohort and same instrument is real signal. Smaller moves sit inside the noise floor.
• Against industry benchmarks. Published SaaS benchmarks put the median NPS around +36, with the strongest performers well above. Treat these vendor-published numbers as rough reference points, not targets — they are self-reported and inconsistently measured.
• Against behavior. Compare each NPS bracket against retention, account expansion, and referral activity in your own data. If your promoters do not actually retain or refer at higher rates than your passives, the score is not telling you much about loyalty.

The open-ended “why” is where the useful insight lives. Group the written answers by score bracket and look for repeated language: what promoters praise tells you what is working; what detractors complain about repeatedly tells you what to fix first.

Segment before you act on the headline. A single company-wide number blurs distinct cohorts. If you have meaningful groups — plan tier, acquisition channel, tenure, or region (kept within consistent geographic groupings, per the cultural-scoring bias below) — recompute the score within each. A +20 overall can hide a +50 enterprise cohort and a -10 free-tier cohort, and those two readings point at opposite actions. Only cut where you have enough responses per group (aim for 30 or more) to trust it.

Mine the Passives, not just the Detractors. Closing the loop on detractors is the canonical NPS move, but Passives (7-8) are usually the cheapest points to gain: they already like the product and sit one or two notches below recommending it. Their open-ended answers tell you the specific gap between satisfied and enthusiastic — often a single missing capability or unresolved friction. Investigating why a 7-8 isn’t a 9-10 gives you a clearer, lower-effort path to lifting the score than winning back detractors who may never have been a fit.

AI analysis tools sort hundreds of written answers into themes and surface the most common reason behind each bracket in minutes. Spend your own attention on the outliers and the ambiguous answers the AI flags rather than reading every response one by one. Then have AI line the scores up against behavior — how often customers use the product, who is at risk of leaving, who refers others — to check that your promoters are actually driving growth.