NPS, CSAT and CES calculator with margin of error

Net Promoter Score (NPS) was introduced by Fred Reichheld in “The One Number You Need to Grow”, Harvard Business Review, 2003. The metric is built on a single question:

Responses fall into three groups: promoters (9–10) — loyal customers who recommend the product; passives (7–8) — satisfied but easily lured by a competitor; detractors (0–6) — unhappy and may talk others out of it. The score:

NPS is measured in points from −100 (all detractors) to +100 (all promoters), not in percent. Passives are not in the numerator but are in the denominator: if promoters turn into passives, the score falls.

Net Promoter®, NPS® and Net Promoter Score® are registered trademarks of Bain & Company, Inc., NICE Satmetrix (Satmetrix Systems, Inc.) and Fred Reichheld. This calculator is an independent tool and is not affiliated with the trademark owners.

NPS is the difference of two shares from the same sample, so its margin of error is larger than that of either share. If p_P and p_D are the shares of promoters and detractors, the variance of the score (in proportions) is:

For the example above (0.6 and 0.167; n = 300), SE = 4.39 and the 95% interval runs from +34.7 to +51.9. A 17-point spread at 300 responses is normal, and it is the key thing to remember when reading reports: an NPS change of 3–5 points with a few hundred responses is almost always within random variation. In the worst case — equal promoters and detractors and no passives — the 95% margin is ±9.8 points at 400 responses; narrowing it to ±5 takes 1,537 responses.

Did NPS really change?

For two independent surveys — say, two quarters with different respondents — test the difference with a z-test:

If the p-value is below 1 − confidence level, the change is significant. If the same people answered both times, the surveys are dependent and the formula is too conservative — panel surveys need paired methods.

CSAT — satisfaction

Customer Satisfaction Score answers “How satisfied are you with…” — usually right after a specific interaction: a support request, a purchase, finishing setup. CSAT has no standard wording; most often it uses a 1–5 scale and counts the share of satisfied customers, i.e. 4 and 5 ratings:

An average rating (“4.2 out of 5”) is also used but is harder to interpret: the same average comes from “everyone gave 4” and from “half gave 5, some gave 1”. For the share of satisfied customers the calculator uses the Wilson interval — at shares around 80–90% the plain “± margin” interval is noticeably off.

CES — customer effort

Customer Effort Score appeared after “Stop Trying to Delight Your Customers” by Matthew Dixon, Karen Freeman and Nicholas Toman (Harvard Business Review, 2010). The authors argued that loyalty depends more on how easy it is to solve a problem than on attempts to “delight”. The first version asked how much effort was required on a 1–5 scale. “The Effortless Experience” (2013) describes the second version: agreement with “The company made it easy for me to handle my issue” on a 1–7 scale, where higher is easier. The calculator uses that version.

Besides the mean, look at the share of 1–3 ratings: these are the people who struggled, and their scenarios deserve attention first.

Which metric to use

• NPS — attitude to the product as a whole. Ask the same user no more than once a quarter, and not before they have got value.
• CSAT — a specific moment: a support ticket, a new feature, a setup step.
• CES — how easy a task is: onboarding, data import, payment. Good at showing where the interface gets in the way.

NPS is popular because it is simple, but the metric has well-known weaknesses.

• The link to growth is disputed. A study by Keiningham and colleagues in the Journal of Marketing (2007) did not confirm that NPS predicts revenue growth better than other satisfaction measures.
• Information loss. Eleven levels collapse into three groups: a rating moving from 0 to 6 does not change the score, but from 6 to 7 it does. Look at the rating distribution, not only the final number.
• Comparing with other companies. NPS levels depend heavily on industry, audience, channel and question wording. We do not quote “benchmarks” — it is more honest to compare a product with itself over time using the same method.
• Self-selection. Not everyone answers, and responders differ from the silent. The margin of error here covers only random sampling error — see the sample size calculator.

When to ask inside the product

The most common mistake is showing NPS on first login. The person has not done anything yet, so the rating reflects the landing page and expectations, not the product. Better:

• ask after a key onboarding step — the first project created, data imported, a teammate invited;
• for CSAT and CES, ask right after the event you are measuring, while the impression is fresh;
• do not interrupt a task: survey after the action is complete, not in the middle of a form;
• always add an open question, “Why did you give this score?” — the number shows what is happening, comments explain why;
• limit frequency: the same person should not see the survey every week.

1. Reichheld F. F. The One Number You Need to Grow. Harvard Business Review, December 2003 — definition of NPS.
2. Dixon M., Freeman K., Toman N. Stop Trying to Delight Your Customers. Harvard Business Review, July–August 2010 — Customer Effort Score.
3. Dixon M., Toman N., DeLisi R. The Effortless Experience: Conquering the New Battleground for Customer Loyalty. Portfolio/Penguin, 2013 — CES 2.0 on a 1–7 scale.
4. Keiningham T. L., Cooil B., Andreassen T. W., Aksoy L. A Longitudinal Examination of Net Promoter and Firm Revenue Growth. Journal of Marketing, 71(3), 2007 — a critique of NPS.
5. Wilson E. B. Probable Inference, the Law of Succession, and Statistical Inference. Journal of the American Statistical Association, 22(158), 1927 — interval for a proportion.
6. ISO 10004:2018. Quality management — Customer satisfaction — Guidelines for monitoring and measuring.