Skip to content

Guide

How many responses do you need?

How many responses you need depends on the confidence level you want and the margin of error you can accept, not on what share of your audience replies. For a large population, the common standard of 95% confidence with a ±5% margin of error calls for about 385 completed responses. If your total population is small and known, you need fewer, and a sample size calculator gives you the exact number for your case.

It is not a percentage of your audience

A common misconception is that you need a fixed slice of your audience, like 10 percent. Statistics does not work that way. Two things decide your sample size: the confidence level, which is how sure you want to be that your result reflects reality, and the margin of error, which is how much wobble around each result you are willing to accept.

Confidence is usually written as a percentage, most often 95 percent, meaning that if you ran the survey many times the true value would fall inside your margin in 95 of them. The margin of error, written as something like ±5 percent, is the band around each result.

The 385 rule of thumb

For a large population, meaning anything from a few thousand people upward where population size stops mattering, the widely used pairing of 95 percent confidence and a ±5 percent margin points to roughly 385 completed responses. That single figure is why so many surveys aim for about 400.

This gets you reliable estimates for the whole group. To pin down the exact figure for a different confidence level or margin, run your numbers through a sample size calculator.

One key caveat: these figures assume your responses come from an unbiased sample. Sample size controls the margin of error, not selection bias: a large but skewed sample does not represent the whole group however many responses it has. Who you invite and who ends up replying matters as much as how many do.

That figure isn't arbitrary: it comes from the classic sample-size formula that Cochran sets out in Sampling Techniques, which at 95% confidence, a 50% proportion, and a ±5% margin gives 384.16, rounded up to 385.

A meta-analysis by Groves and Peytcheva published in 2008 in Public Opinion Quarterly, which pooled 59 studies with real bias measures, found that response rate is a poor predictor of how much nonresponse bias a survey carries.

Small, known populations need fewer

If you are surveying a group you can count, say 200 employees or 80 suppliers, you need far fewer than 385. A statistical adjustment called the finite population correction lowers the target as the population shrinks.

To reach the same 95 percent and ±5 percent standard, a population of 200 needs around 130 responses rather than 385. The smaller and better defined your audience, the closer your required sample gets to simply asking most of them.

The Krejcie and Morgan reference table (1970), still in common use today, sets that requirement precisely: for a population of 200 people, 132 responses.

More responses, a tighter margin

Responses and precision are linked, but not one to one. Adding responses narrows the margin of error with diminishing returns: roughly speaking, halving your margin quadruples the responses you need. Moving from ±5 percent to ±2.5 percent is a large jump in effort for a modest gain.

Decide what precision the decision actually needs before chasing more responses. For a directional read on whether people like an idea, a wide margin is usually fine. For a close call between two options, you may want a tighter one.

Watch the count on each question and segment

Your total sample only protects estimates for the whole group. The moment you filter down to, say, women under 30 or a single region, the count behind that slice drops and its margin of error widens.

When a segment or a single question rests on only a handful of responses, treat the result as directional: a hint worth exploring, not a firm number to report. If a breakdown matters to you, plan enough responses so every cell you care about keeps a usable count.

As answers arrive, Camaleonic Survey shows results live, which makes it easy to see when each segment has reached a count you can trust.

Frequently asked questions

Is 385 responses always enough?
It is the standard target for a large population at 95 percent confidence with a ±5 percent margin, and it is enough for reliable results about the whole group. It is not enough for detailed breakdowns, because once you split the data into segments each slice needs its own healthy count. If you want a margin tighter than ±5 percent, you will need more.
What if only a few people respond?
Fewer responses do not make a survey worthless, they simply widen the margin of error. With a small sample, read the results as directional signals rather than precise figures, and be careful about acting on small differences between options. Always state your sample size when you share the findings so others can weigh them fairly.
Do I need a bigger sample if my audience is huge?
No. Beyond a few thousand people, population size barely changes the sample you need, which is why the same figure of about 385 works for a city or a whole country at 95 percent and ±5 percent. What raises the number is a higher confidence level, a tighter margin, or wanting reliable results for many separate segments.

Sources

  1. Groves, R. M. & Peytcheva, E. (2008). The Impact of Nonresponse Rates on Nonresponse Bias: A Meta-Analysis. Public Opinion Quarterly, 72(2), 167-189The meta-analysis that separated response rate from the bias a survey actually carries.
  2. Cochran, W. G. (1977). Sampling Techniques, 3rd ed. WileyThe sample size formula and the finite population correction our calculators implement.
  3. Krejcie, R. V. & Morgan, D. W. (1970). Determining Sample Size for Research ActivitiesThe classic sample size table by population, which yields the same figures as the formula.

Build your survey with Camaleonic Survey

AI to draft it, your brand on every screen, and live results as responses arrive.

Start free