Calculator · Survey planning
Sample size calculator: how many responses does your survey need?
How many responses does your survey need? This sample size calculator works it out from the confidence level, margin of error, expected proportion and – if you know it – the size of the population. It also shows how many people to invite at your response rate and, the other way round, how precise an existing number of responses is.
Building and testing is free. To publish, you choose a paid plan.
Sample size calculator
The starting values are the usual choice: 95% confidence level, ±5% margin of error and a 50% expected proportion. Everything is calculated in your browser, and nothing is stored.
Responses needed
385
Invitations needed at a 20% response rate: 1,925
Calculated with a 95% confidence level, ±5% margin of error and 50% expected proportion: n = z² · p · (1 − p) / e², with the finite population correction if you entered a population, rounded up.
The other way round: how precise are the responses you have?
Margin of error for this number of responses (200): ±6.9%
Sample size explained
What does the confidence level mean?
The confidence level says how sure you want to be that the true value lies within your margin of error. 95% means: if you repeated the same survey many times with new random samples, the true value would fall within the calculated range in about 95 out of 100 cases. More certainty costs responses: at ±5% and a 50% proportion, you need 271 responses for a 90% confidence level, 385 for 95% and as many as 664 for 99%. In market and social research, 95% is the usual standard.
What is the margin of error?
The margin of error is half the width of the confidence interval: it tells you how far your sample result may be from the true value in the population. If 60% of respondents say yes and the margin of error is ±5%, the true share probably lies between 55% and 65%. The margin of error only shrinks with the square root of the sample size: to halve it, you need four times as many responses. 385 responses at ±5% become 1,068 at ±3%. And if you analyse subgroups, each group has its own, larger margin of error.
Why is 50% the default proportion?
The formula contains the product p · (1 − p), which is largest at 50%. Using 50% therefore means planning for the worst case, so you are on the safe side whatever the answers turn out to be. If you know from earlier surveys that around 10% give a particular answer, you can work with 10% – then 139 responses are enough at 95% and ±5%. If your survey asks about many different things, stick with 50%, because the sample has to be large enough for the least favourable question, not just one of them.
When does the population size matter?
Hardly at all for large groups: whether it is 100,000 or ten million people, the sample you need at 95% and ±5% stays almost the same. For small populations, however, the finite population correction lowers the number considerably: for 1,000 people, 278 responses are enough; for 500 it is 218, and for 100 people it is 80. The formula is n = n₀ / (1 + (n₀ − 1) / N), where n₀ is the sample for an unlimited population. For small groups such as a company with 100 employees, surveying everyone often makes sense – you would be inviting almost all of them anyway.
How do I plan around the response rate?
The calculator gives you the number of complete responses, not the number of invitations. If experience tells you that one in five invited people responds, i.e. 20%, you need 1,925 invitations for 385 responses. Estimate the rate cautiously, ideally from previous surveys of the same audience. Important: a large sample does not make up for bias. If it is mainly the satisfied customers or a single age group who respond, even a small margin of error will not help. With quotas in 121surveys you set how many responses you need per group; once a group is full, the survey ends for further participants from that group.
Frequently asked questions
Is the sample size calculator free?
Yes, with no account and no sign-up. All values are calculated in your browser and are neither stored nor sent anywhere.
How many responses do I need for a representative survey?
For a 95% confidence level and a ±5% margin of error in a large population, it is 385 responses. But a survey does not become representative through numbers alone: the respondents also have to reflect the make-up of the population. A random selection or quotas for key characteristics matter more than any extra response.
Does the formula also work for averages, such as ratings from 1 to 5?
The calculator works with proportions, i.e. questions like “What percentage agree?”. For averages you need an estimate of the spread (standard deviation) and a different formula. The figures for proportions are still a useful rough guide, because many analyses end up reporting proportions – such as the share of top-two ratings.
What does the margin of error mean for subgroups?
The margin of error applies to the group you want to say something about. If you want to report results separately for three regions, each region needs a large enough sample of its own. The reverse calculation in the calculator shows how precise a subgroup is with its number of responses: with 200 responses the margin of error is around ±6.9%.
Can I still use a small sample?
Yes, as long as you report the uncertainty alongside the results. Fewer responses are fine for early indications, qualitative insights or testing a survey. For decisions that hinge on a few percentage points, however, you need a small margin of error – and therefore many responses.
From sample size to finished survey
Create your survey in 121surveys, control responses per group with quotas and try everything for free with the test link.
