Net Promoter Score (NPS)

Respondents are split into three groups: promoters who answer 9 or 10, passives who answer 7 or 8, and detractors who answer 0 to 6. Passives are counted in the base but not in the score, so NPS can range from negative 100 to positive 100. The number is a snapshot of stated intent to recommend, which is a proxy for word of mouth and a rough gauge of how a customer base feels at one point in time.

The survey is most honest when the timing matches the experience. A relational NPS asks the whole customer base on a fixed cadence, say once a quarter, and tracks broad sentiment. A transactional NPS fires after a specific moment, a delivery, a return, a support reply, and tells you whether that moment helped or hurt. Mixing the two in one figure muddies the signal, so most teams keep them on separate dashboards and read them separately.

Consider a Shopify store selling cycling apparel. It sends a transactional survey five days after delivery and reads 200 responses: 110 promoters, 50 passives, 40 detractors. That works out to 55 percent promoters minus 20 percent detractors, an NPS of 35. The digit is fine, but the open comments are where the work is. A cluster of detractors all mention that a jersey ran a size small. The fix is a sizing note on the product page and a revised size chart, not a campaign to lift the score.

What NPS does not tell you on its own is why. A single number hides the reasons behind it, so the follow-up comment is usually more useful than the digit. The metric is also sensitive to sampling, timing, and culture, since scoring norms differ by region, so treat absolute values with care and watch the trend over time rather than chasing a published benchmark.

NPS and reviews measure the same underlying goodwill from two angles: one is a private survey number, the other is public, searchable testimony. This matters for AI search and answer engines. When a shopper asks ChatGPT, Perplexity, or Google AI Overviews whether a brand is worth buying, those systems read what is visible: review text, ratings, and the language customers use, not a loyalty score locked in your analytics. A high NPS that never becomes a published review does little at the moment a model is deciding which brands to name. The practical move is to route your promoters toward leaving a review, so private goodwill turns into the public evidence answer engines can actually cite.