Customer experience metrics
CSAT and CES calculator
How results are checked
Named statistical review pending. Selected automated reference cases are published on the validation page.CSAT, top-2-box (4–5 of 5)
63.1%satisfied
95% Wilson CI 58.7% to 67.4%
Mean 3.83 of 5 · CES 7-point: 68.8% agree it was easy (5–7) · n = 472
CSAT is the percentage of respondents in the top boxes of a satisfaction scale — usually 4 or 5 on 1–5 (“top-2-box”). CES, the Customer Effort Score, is either the share agreeing (5–7) that “the company made it easy for me to handle my issue” on a 7-point scale, where higher is better, or the mean of the older 1–5 effort question, where lower is better. This calculator gives both the share (with a Wilson interval) and the mean (with a t interval), tests two groups, and breaks results down by segment. Free, no account, and files are never uploaded.
Calculate CSAT and CES online
Pick the metric and scale, then enter the number of responses at each score, compare two groups, or open a survey file and choose the rating question. Everything is calculated in this browser tab.
CSAT: the share of respondents in the top boxes of the 1–5 satisfaction scale (4–5) and the mean score. Higher is better.
Number of responses at each score (1 = very dissatisfied, 5 = very satisfied).
Satisfied (4–5)
76.7%
Satisfied (4–5): 76.7% (95% Wilson CI 71.6% to 81.1%); Mean satisfaction 4.08 (95% CI 3.95 to 4.20).
95% CI 71.6% to 81.1% · n = 300
| Satisfied (4–5) — 95% CI (Wilson score) | 76.7% · 71.6% to 81.1% |
| Mean satisfaction — 95% CI (t, 299 df) | 4.08 · 3.95 to 4.20 |
| Standard deviation / standard error of the mean | 1.087 / 0.063 |
Number of responses at each score (1 = very dissatisfied, 5 = very satisfied).
Difference in Satisfied (4–5) (B − A)
−10.0 pts
Favourable share: −10.0 percentage points, significant at the 95% level, p = .007 (two-proportion z-test). Mean: −0.26, significant at the 95% level, p = .006 (Welch t-test).
z = -2.72 · 95% CI −17.1 to −2.8 pts
| n | Satisfied (4–5) | 95% CI | Mean | 95% CI | |
|---|---|---|---|---|---|
| Group A | 300 | 76.7% | 71.6% to 81.1% | 4.08 | 3.95 to 4.20 |
| Group B | 300 | 66.7% | 61.2% to 71.8% | 3.82 | 3.68 to 3.95 |
| Difference (B − A) | — | −10.0 pts | −17.1 to −2.8 | −0.26 | −0.44 to −0.08 |
Share: pooled two-proportion z-test, Newcombe hybrid score interval for the difference. Mean: Welch's unequal-variance t-test (t = -2.77, df = 591.6). Both assume independent groups.
CSV, Excel, SPSS, Stata, SAS, R, Parquet, SQLite or JSON. The file is read in this browser tab and never uploaded. For Triple-S, select the .sss or .xml metadata file and its data file together.
Free, no account needed. Counts and files stay in this browser tab and are never uploaded.
Which CSAT and CES version you have
CSAT has no single standard scale. 1–5 is the most common, with 1–7 and 1–10 also in use; the tool lets you choose the scale and how many top boxes count as satisfied. Always report which boxes you counted (“top-2-box, 4–5 of 5”), because top-1 and top-2 shares are not comparable.
CES exists in two versions with opposite directions. The 7-point agreement item (“The company made it easy for me to handle my issue”, 1 = strongly disagree to 7 = strongly agree) was introduced by Dixon, Toman and DeLisi (2013); higher is better, and the usual summary is the share agreeing (5–7) or the mean. The original item (Dixon, Freeman and Toman 2010) asked “How much effort did you personally have to put forth to handle your request?” on a 1–5 scale from very low to very high effort, so lower is better. The tool labels the direction everywhere and treats the low-effort boxes (1–2) as favourable on that scale.
For likelihood to recommend on a 0–10 scale, use the NPS calculator instead.
Is there a good CSAT or CES score?
This page does not quote a “good” CSAT or CES. Published benchmarks depend on the industry, the touchpoint being rated, the exact wording and scale, the survey mode and who gets invited, so a number from another company's survey is rarely comparable with yours. The more reliable reference is your own trend: track the same question, scale and box definition wave after wave, and use the two-group test (or the confidence interval) to tell real movement from sampling noise.
The method
Formula
CSAT = x / n, Wilson CI = [p + z²/2n ± z·√(p(1 − p)/n + z²/4n²)] / (1 + z²/n), mean CI = x̄ ± t(n − 1) · s / √n
where
- x
- responses in the favourable boxes (e.g. 4–5 of 5; 5–7 on the 7-point CES; 1–2 on the 1–5 effort scale)
- n
- valid responses (Kish effective n = (Σw)²/Σw² for weighted data)
- p
- favourable share x / n
- z
- 1.645, 1.960 or 2.576 for 90%, 95% or 99% confidence
- x̄, s
- mean score and its standard deviation (n − 1 divisor)
- Two groups
- share: pooled z = (p_B − p_A) / √[p̄(1 − p̄)(1/n_A + 1/n_B)], Newcombe interval for the difference; mean: Welch t = (x̄_B − x̄_A) / √(s_A²/n_A + s_B²/n_B) with Welch–Satterthwaite df
Worked example
The synthetic customer-support survey loaded by “Use sample data” has 480 respondents; 472 answered the 1–5 satisfaction question (8 left it blank). 298 chose 4 or 5, so CSAT (top-2-box) = 298 / 472 = 63.1%.
The Wilson interval centres on (p + z²/2n) / (1 + z²/n) with z² = 3.841, which pulls it slightly towards 50%: the 95% interval is 58.7% to 67.4%. The mean score is 3.828 with SD 0.880, so SE = 0.880 / √472 = 0.0405 and the t interval on 471 df is 3.75 to 3.91. On the 7-point effort item, 68.8% agree (5–7) that the company made it easy (mean 5.15).
By support channel, CSAT is Chat 78.8% (n = 132); Email 48.9% (n = 94); Phone 59.4% (n = 160); Self-service 61.6% (n = 86). Chat (+21.7 points vs the rest) and Email (−17.7 points vs the rest) differ from the other respondents at the 95% level; the other channels are within sampling noise. These are synthetic numbers, not a benchmark.
Using it well
When to use it
Use it when
- Use: You have answers to a single satisfaction or effort rating question on a 1–5, 1–7 or 1–10 scale.
- Use: You need CSAT or CES with a margin of error, or want to know whether two waves, channels or teams really differ.
- Use: You want the top-box share and the mean side by side, by segment, from a respondent-level file.
Not the right tool when
- Avoid: The two groups contain the same respondents (a panel asked twice) — the independent-samples tests do not apply.
- Avoid: You want to compare your score with another company's published figure — wording, scale and sample differ; track your own trend instead.
- Avoid: You need to know what drives satisfaction — use key driver analysis on the attribute ratings.
How to interpret it
Rule of thumb
If the confidence interval of a difference excludes zero (equivalently p < 0.05 at 95%), the two groups differ beyond sampling error. The share and the mean can disagree: a group can gain top-box respondents while its mean barely moves. Report the box definition and the scale with every number, and on the 1–5 effort scale remember that a falling mean is an improvement. Weighted samples should be weighted before scoring; the tool then uses the effective sample size.
Questions
Frequently asked questions
How is CSAT calculated?
Count the respondents who chose a top box on the satisfaction scale — usually 4 or 5 on a 1–5 scale — and divide by everyone who answered. For example, 230 satisfied out of 300 answers gives CSAT = 76.7%. Say which boxes you counted when you report it.
How is the Customer Effort Score calculated?
On the 7-point version, take the share who agree (5, 6 or 7) that the company made it easy to handle their issue, or the mean score; higher is better. On the older 1–5 version, which asks how much effort the customer had to put in, report the mean (lower is better) or the share choosing 1–2 (low effort).
Why a Wilson interval instead of p ± 1.96·SE?
CSAT shares are often close to 100%, where the simple Wald interval can run past 100% and covers the true value less often than stated. The Wilson score interval stays inside 0–100% and keeps close to the nominal coverage even for small samples or extreme shares.
How do I test whether two CSAT scores differ significantly?
For two independent groups, the 'Compare two groups' tab runs a two-proportion z-test on the top-box shares and Welch's t-test on the means, and reports p-values and the confidence interval of each difference.
Is my survey file uploaded?
No. The file is parsed in this browser tab, the calculations run locally, and the CSV export is created on your device. No account is needed.
Sources and next steps
References & further reading
- [1]Wilson, E. B. (1927). Probable inference, the law of succession, and statistical inference. Journal of the American Statistical Association, 22(158), 209–212.
- [2]Newcombe, R. G. (1998). Interval estimation for the difference between independent proportions: comparison of eleven methods. Statistics in Medicine, 17(8), 873–890.
- [3]Welch, B. L. (1947). The generalization of ‘Student’s’ problem when several different population variances are involved. Biometrika, 34(1–2), 28–35.
- [4]Dixon, M., Freeman, K., & Toman, N. (2010). Stop trying to delight your customers. Harvard Business Review, 88(7/8), 116–122.
- [5]Dixon, M., Toman, N., & DeLisi, R. (2013). The Effortless Experience. New York: Portfolio/Penguin.
- [6]Kish, L. (1965). Survey Sampling. New York: Wiley.
Try it on your own data — free, no signup
Upload a CSV, Excel, SPSS, Stata, SAS, R, Parquet or JSON file. Everything runs in your browser; your file never leaves your device.
Open the workspace →