What Is Sample Size?
The number of respondents a survey needs for its results to reliably represent the larger population being studied.
Sample size is how many completed responses you need before you can trust the results as representative of a bigger group — all your employees, all your customers, a target market — rather than just an opinion from whoever happened to answer.
Required sample size is driven by three things: how large the total population is, how tight a margin of error you’re willing to accept, and how confident you want to be in the result (typically expressed as a 95% or 99% confidence level). Counterintuitively, once a population gets large enough (a few tens of thousands and up), the required sample size barely grows any further — surveying a population of 50,000 and a population of 5 million require a very similar sample size for the same margin of error.
Undersized samples produce results that swing wildly on small changes in who happened to respond; oversized samples cost more time and incentive budget without meaningfully tightening the result. The right size is the smallest one that gets you the margin of error and confidence level your decision actually needs.
Example
| Margin of error | Responses needed |
|---|---|
| ±10% | 97 |
| ±5% | 385 |
| ±3% | 1,068 |
- Where 385 comes from (±5%)
- 1.96² × 0.5 × 0.5 ÷ 0.05²=385The sum comes to 384.16, rounded up. 1.96 is the figure for 95% confidence, and 0.5 assumes answers split evenly, the cautious case. A small population needs fewer.
±5% at 95% confidence is the most common "good enough for most business decisions" target.
What it does not tell you
- The figures assume a simple random sample. If people chose themselves to respond, or you only reached whoever was convenient, the formula describes a precision you do not actually have.
- It sizes the whole sample. Results for a subgroup, such as one region or one team, rest on that subgroup's own count and carry a wider margin of error.
- It estimates a share, such as the percentage who agree. Comparing two groups, or detecting a change over time, needs a power calculation and usually more responses.
- Clustered or stratified designs depart from simple random sampling and need an adjustment known as the design effect.
Sources
- Cochran, W. G. (1977). Sampling Techniques (3rd ed.). John Wiley & Sons.
How to Calculate Sample Size for SurveyRock
Use SurveyRock's free sample size calculator to work out exactly how many responses you need for your population, margin of error, and confidence level before you launch.