🔬 Sample Size Calculator
Set your confidence level, margin of error, and expected proportion to see how many responses your survey needs — with an optional population size for the finite-population correction.
🔬 Size Your Study
What is a Sample Size Calculator?
It tells you how many observations you need to collect so that your estimate of a proportion is precise and reliable. Give it the confidence level you want, the margin of error you can tolerate, and your best guess at the proportion, and it applies Cochran's formula to return the minimum sample — optionally corrected for a known, finite population.
Plan surveys, A/B tests, and quality audits with it before you spend time and budget collecting data. Too small a sample yields wide, untrustworthy intervals; an unnecessarily large one wastes resources. This gets you to the right number for the precision you actually need.
❓ Frequently Asked Questions
How is the required sample size calculated?
This tool uses Cochran's formula, n₀ = (z² · p · (1 − p)) / e², where z is the z-score for your confidence level, p is the expected proportion, and e is the margin of error. The result is rounded up to a whole number of respondents. When you supply a finite population, a correction is applied to reduce that figure.
Which confidence level and margin of error should I choose?
A 95% confidence level with a 5% margin of error is the common default for surveys. Higher confidence (99%) or a tighter margin (3%) both increase the sample you need. Confidence is how often the true value would fall within your interval across repeated samples; margin of error is the ± band around your estimate.
Why does a proportion of 50% require the largest sample?
The term p · (1 − p) is maximised at p = 0.5, so assuming a 50% proportion gives the most conservative — largest — sample size. If you genuinely have no prior estimate of the proportion, use 50% to be safe; if you have a reliable estimate, using it will lower the required sample.
What does the finite-population correction do?
When you are sampling from a limited population rather than an effectively infinite one, you don't need as many responses. The correction n = n₀ / (1 + (n₀ − 1)/N) shrinks the requirement toward the population size N. For very large populations the corrected figure barely differs from the infinite-population result.