Survey sample size calculator
How many responses you need for a result you can rely on. It names the formula and the assumptions it used, so the number is something you can defend rather than something you have to trust.
To be 95% confident your result is within ±5% of the real answer, from a population of 1,000.
Cochran's formula with z = 1.96, p = 0.5, giving 384.16 before any population adjustment, then corrected for a finite population. p is held at 0.5 because that is the response split needing the largest sample, so the number is never optimistic.
Once you know the target, SurveyService collects the responses and writes up what they say. See how it works →
How the number is worked out
This uses Cochran's formula, the standard basis for sample size. It first works out how many responses an unlimited population would need, then adjusts that down for the fact that yours is finite: surveying 80 of 100 people tells you much more than 80 of a million.
Three inputs set the answer. Population is how many people could answer at all. Confidence level is how sure you want to be, with 95% the standard default. Margin of error is how much slack you will accept in the result, with ±5% standard. The response split is held at 50/50 throughout, because that is the split requiring the most responses, so the number this returns is never optimistic.
Responses needed by population size
At 95% confidence and a ±5% margin of error. Note how little it changes once the population passes a few thousand, which is why national polls and global surveys use numbers that look surprisingly small.
| Population size | Responses needed |
|---|---|
| 100 | 80 |
| 500 | 218 |
| 1,000 | 278 |
| 5,000 | 357 |
| 10,000 | 370 |
| 100,000+ | 385 |
Frequently asked questions
How many responses does a survey need?
For a large or unknown population at the standard 95% confidence level and a 5% margin of error, about 385. That number barely moves once the population is large, so 385 is a reasonable target for most public or large-audience surveys. Smaller populations need far fewer: 278 for a thousand people, 80 for a hundred.
What do confidence level and margin of error actually mean?
Confidence level is how often the method would land close to the truth if you repeated the survey: 95% is the standard default. Margin of error is how far the result could sit from the real answer, so a ±5% margin on a 60% result means the truth is likely between 55% and 65%. Tightening either one costs responses, and tightening the margin costs far more than raising confidence.
Why does the calculator assume a 50/50 response split?
Because that split needs the largest sample, so it is the conservative choice. Assuming a more lopsided split would return a smaller number, which looks appealing until the responses come back closer to even and the margin of error you planned for turns out to be wider than you thought.
What if I cannot get that many responses?
Run the survey anyway and be honest about the margin. Fewer responses does not make a survey worthless, it makes it less precise, and knowing roughly what people think is usually better than knowing nothing. What matters is not treating a small sample as though it were a large one when you report it.
More on this: the full guide to survey sample size. Already have responses? Try the NPS calculator.