NPS asks one question — "how likely are you to recommend us?" (0–10) — and nets promoters (9–10) against detractors (0–6). It's a directional pulse on sentiment, not a precise metric.
Example: 60% promoters − 15% detractors = an NPS of 45. Track the trend and read the verbatim comments; the number alone tells you little about why.
Above 0 is good; above 50 is excellent.
Example — NPS in practice
Picture Talabat surveying 1,000 customers after iftar-time deliveries during Ramadan, asking how likely they are to recommend the app. 620 respondents are promoters (scoring 9–10), 260 are passives, and 120 are detractors. NPS is promoters minus detractors: 62% − 12% = 50, a strong score suggesting delivery speed during the rush is winning loyalty rather than eroding it.
نفترض أن طلبات تستطلع آراء 1000 عميل بعد توصيل وجبات الإفطار في رمضان، وتسألهم عن مدى استعدادهم للتوصية بالتطبيق. من بينهم 620 من المروّجين (بتقييم 9-10)، و260 محايدين، و120 من المنتقدين. صافي نقاط الترويج يُحسب بطرح نسبة المنتقدين من نسبة المروّجين: 62% - 12% = 50، وهي نتيجة قوية تشير إلى أن سرعة التوصيل في ساعة الذروة تعزز الولاء بدلاً من أن تضعفه.
NPS, properly understood
NPS comes from a single question — 'How likely are you to recommend [product] to a friend or colleague?' — scored 0-10. Respondents scoring 9-10 are Promoters, 7-8 are Passives (counted in the denominator but not the formula), and 0-6 are Detractors. The score is %Promoters minus %Detractors, expressed as a plain integer from -100 to +100, not a percentage. Data sources are typically an in-app or post-purchase survey (a modal, an SMS/WhatsApp follow-up, or a custom tool) sent within a day or two of a key moment — delivery, support resolution, renewal — because recall decays fast and a survey sent weeks later measures general brand sentiment, not the triggering event.
Timing and channel matter more in the GCC than the raw script. A WhatsApp-delivered NPS survey after a COD delivery typically gets far higher response rates than email, but respondents skew toward people who just had the transaction top-of-mind, which can inflate the score relative to a random-sample survey. Ramadan and Eid delivery windows are notoriously bad for logistics-heavy businesses — asking for a score during peak congestion captures operational strain rather than product sentiment, so many operators run NPS on a rolling basis and flag Ramadan-window responses separately rather than blending them into the annual trend. Bilingual surveys need genuinely separate open-text analysis: an Arabic 'ممتاز' and an English 'fine' can both be a 9, but the qualitative signal underneath is not the same.
NPS is a lagging, low-frequency signal masquerading as a real-time one — a single quarterly number hides which specific moment (onboarding, support, pricing) is driving detractors, so it should always ship with a follow-up 'why' question. Small sample sizes swing wildly: a handful of ratings can move the score 20 points, so treat anything under a few hundred responses as directional. Comparing NPS across industries or countries is close to meaningless without a like-for-like benchmark, since response culture varies (some markets rate generously, others conservatively), and comparing your own score against an unverified public 'industry average' invites false confidence either way.
Pair NPS with CSAT (a per-interaction satisfaction score, better for diagnosing a single touchpoint) and with a genuine retention or churn number, since the real test of NPS is whether it predicts renewal and referral behavior, not whether it looks good on a slide. Track the open-text verbatims as a qualitative feed into support and product backlogs — the score without the comments is close to useless for actually fixing anything.
Put it to work
- Trigger the survey within 24-48 hours of the moment being measured, not weeks later.
- Always send it with an open-text 'why' follow-up — the score alone doesn't tell you what to fix.
- Segment Ramadan/Eid-window responses separately from your baseline trend for logistics-heavy products.
- Treat scores from under a few hundred responses as directional, not decision-grade.
- Route verbatims to support and product as a standing backlog input, not just a quarterly report.
- Run Arabic and English verbatims through separate qualitative review — a matching score can hide different sentiment.
- Localize the survey question and scale for Arabic rather than machine-translating it directly — stiff or overly formal phrasing measurably depresses both response rate and the willingness to select an extreme score.
Turn the theory into real pipeline.
Get a free 60-second growth audit of your site, or talk to a strategist about your funnel.