Find out how many responses you need before your employee survey result means anything — and what response rate that implies for your headcount.
Your survey
152
responses needed from 250 people — a 61% response rate
95%
Confidence level
±5%
Margin of error
61%
Response rate required
A normal, healthy target for a well-run survey with reminders and leadership backing.
Calculated with the standard finite-population formula at maximum variance (p = 0.5), which is the safe assumption when you do not know how opinion will split. Nothing is sent anywhere.
Employee surveys have to clear two separate bars, and only one of them is statistical.
This is what the calculator above gives you: enough responses that the number you report is close to the number you would have got from everyone. It scales sub-linearly, which is why a 1,000-person company needs a far lower response rate than a 50-person one.
Never report a segment with fewer than five responses, regardless of what the statistics permit, and never cross-tabulate two small segments — “senior engineers in the Berlin office” can easily resolve to one person. The moment employees believe a result could be traced back to them, your next survey gets the answers people think are safe rather than the ones that are true. TruePulse enforces this suppression automatically.
Response rate is a trust metric more than a logistics one. The organisations that consistently clear 80% are the ones that publish results — including the unflattering parts — and name what changed as a result of the last wave.
It depends on headcount. A 50-person company needs about 44 responses for a ±5% margin at 95% confidence — essentially everyone. A 1,000-person company needs about 278, which is a 28% response rate. Smaller organisations always need a proportionally higher response rate, which is why chasing participation matters more the smaller you are.
Margin of error is how far your result could sit from the truth: a 70% score with ±5% means the real figure is somewhere between 65% and 75%. Confidence level is how often that range would contain the truth if you repeated the survey — 95% is the standard choice.
Because it is the most cautious assumption. Variance is highest when opinion divides evenly, so planning for p = 0.5 guarantees your sample is large enough regardless of how the answers actually land.
Yes, and this is where most survey analysis goes wrong. If you plan to report on a 20-person department, that department needs its own adequate sample — a company-wide response rate tells you nothing about whether a specific team's result is meaningful. Separately, never report any segment below five responses, no matter what the statistics say, because anonymity fails before significance does.
Widen the margin of error and say so when you present the results. A survey with 40% participation and an honest caveat is far more useful than one with no caveat and false precision. Then work on the response rate — it is almost always a trust and follow-through problem rather than a survey design problem.
Automatic reminders, smart send timing, and anonymity thresholds enforced by default — so you hit the number without chasing people by hand.
Get started free