Enter how many respondents scored 9 or 10, 7 or 8, and 0 to 6. You get Net Promoter Score, the share in each group, and the margin of error, which on most surveys is larger than the change people are celebrating.
The score runs from minus 100, where everyone is a detractor, to plus 100, where everyone is a promoter. Passives are counted in the total but not in either group, which means converting a passive into a promoter moves the score, and so does converting a passive into a detractor, in the opposite direction.
NPS is built on a single question: how likely are you to recommend this to a friend or colleague, on a scale of zero to ten. Its popularity comes from being short, comparable over time, and easy to explain in a meeting.
What it measures well is broad sentiment within a stable customer base asked the same way each time. What it measures poorly is almost everything else. It does not tell you why people feel as they do, it compresses eleven possible answers into three buckets and discards the difference between a 0 and a 6, and it responds to who you ask and when as much as to the product.
The follow-up question, asking why they gave the score, is where the useful information lives. The number tells you whether to worry. The comments tell you what to fix.
NPS is a difference between two proportions, and differences between proportions are noisy. The worked example, 412 responses scoring +21.4, has a 95 percent margin of error of about 7.7 points. The honest statement is that the true score is somewhere between +13.6 and +29.1.
| Responses | Margin of error | Smallest change worth reacting to |
|---|---|---|
| 50 | ±22 | Almost none |
| 100 | ±16 | Very large swings only |
| 250 | ±10 | About 14 points |
| 412 | ±8 | About 11 points |
| 1,000 | ±5 | About 7 points |
A team presenting a rise from +18 to +24 on 300 responses a quarter is presenting noise. That is not a reason to stop measuring, it is a reason to report the range alongside the score, and to wait for a trend across several periods before claiming a result.
The reason NPS matters to anyone running paid media is referral. Promoters bring in customers who were never paid for, and those customers lower your blended acquisition cost without appearing in any ad platform.
That effect is real but hard to size, and it is frequently overstated. Most promoters never refer anyone. The meaningful link runs through retention rather than referral: detractors churn faster, and churn is what sets the ceiling on what a customer is worth. The retention rate calculator and LTV calculator show why a few points of churn matter so much.
If you want to test whether NPS actually predicts behaviour in your business, compare the repurchase or renewal rate of promoters against detractors a few months after the survey. If the gap is small, the score is measuring mood rather than intent.
On 412 responses, a score of +21 means the truth is somewhere between +14 and +29.
The first trap is survey timing. Asking straight after a delivery arrives captures relief; asking after a support ticket captures frustration. Change when you ask and the score changes without anything about the customer experience moving.
The second is who answers. Response rates on NPS surveys are often below twenty percent, and the people who respond are disproportionately the delighted and the furious. The score describes the responders, not the customer base.
The third is comparing across industries. Published NPS averages vary enormously by sector and by survey method, and a figure from a different industry, channel or question wording tells you almost nothing about yours.
The last is making it a target. Once teams are measured on NPS, they start asking for high scores, surveying the customers most likely to give them, and timing requests around good moments. The number improves and stops meaning anything.
Subtract the percentage of detractors, who scored 0 to 6, from the percentage of promoters, who scored 9 or 10. Passives who scored 7 or 8 count toward the total but not toward either group.
Anything above zero means more promoters than detractors. Beyond that it depends heavily on industry, survey timing and method, so compare against your own previous scores rather than published averages.
More than most teams collect. With 400 responses the margin of error is around 8 points; you need roughly 1,000 to get it down to 5.
Usually sampling noise, a change in who responded, or a change in when you asked. On small samples a swing of ten points or more can happen by chance alone.
We use Google Analytics to count visits, only if you allow it. Nothing you type into a calculator is ever sent. Privacy policy