Sample size calculator
Enter your confidence level and margin of error to find how many people you need to ask.
About Sample Size Calculator
How many responses are enough? This calculator answers with Cochran's formula, the standard starting point for surveys. For a yes or no question it uses the expected proportion, and for a measured quantity it uses the standard deviation. You pick the confidence level between 80 and 99.9 percent, and the matching z-score is worked out exactly rather than looked up from a short table. If you know the size of the population, a finite population correction brings the number down. The working is shown, and the assumptions are listed beside the result.
- Planning a survey for a class project or a thesis.
- Checking whether a sample is large enough.
- Seeing how margin of error changes the sample you need.
- Showing the formula in a methods section.
How to use Sample Size Calculator
Step 1
Choose proportion or mean and set the confidence level.
Step 2
Enter the margin of error and either the expected proportion or the standard deviation.
Step 3
Add the population size if it is small and known, then read the sample size.
Good to know
- If you have no idea what the proportion will be, use 50 percent. It gives the largest sample and is the safe choice.
- The result is the number of completed responses. If only a fraction of the people you contact will answer, divide by the response rate.
- The formula assumes a simple random sample. Clustered or stratified designs need a design-effect adjustment that this tool does not apply.
Common questions
What sample size gives a 5 percent margin of error at 95 percent confidence?
385 for a large population, using p = 50 percent: 1.96 squared x 0.25 / 0.05 squared = 384.16, rounded up.
When should I use the population correction?
When the sample would be more than about 5 percent of the population. For 1,000 people the 385 drops to 278.
Why is the sample size always rounded up?
Rounding down would leave you slightly below the margin of error you wanted, so the result is rounded up to the next whole person.