Survey sample size: how many responses do you actually need?
At the standard 95% confidence level with a ±5% margin of error, you need about 80 responses from a population of 100, 278 from 1,000, 370 from 10,000, and 383 from 100,000 or more. Sample size grows very slowly past a few thousand people — you almost never need more than ~400 responses for ±5% precision, no matter how large the population. What you do need is a sample that actually represents your audience.
What do confidence level and margin of error mean?
Two numbers define how precise a survey result is, and both have plain-English meanings.
Margin of error is the 'plus or minus' on your result. If 60% of respondents say yes with a ±5% margin of error, the true figure in the full population is likely somewhere between 55% and 65%.
Confidence level is how often that range would contain the truth if you repeated the survey many times. At 95% confidence — the standard choice — about 19 out of 20 repeated surveys would produce a range containing the true value. Higher confidence or a tighter margin both require more responses.
How many responses do you need? (95% confidence, ±5%)
| Population size | Responses needed |
|---|---|
| 100 | 80 |
| 500 | 218 |
| 1,000 | 278 |
| 10,000 | 370 |
| 100,000 or more | 383 |
Read the table from your population — everyone your survey is about, such as all customers, all employees, or all event attendees — to the number of completed responses you need. Two things stand out. For small populations you need a surprisingly large share (80 of 100 is 80%). And past about 10,000 people the requirement barely moves: a city of 100,000 and a country of 100 million both need roughly 383 responses for ±5% precision.
Why does sample size stop growing?
The standard formula behind the table is n = z²·p(1−p) ÷ e², where z reflects the confidence level (1.96 for 95%), p is the expected proportion (0.5 is used as the safest, most conservative assumption), and e is the margin of error (0.05). Plugging in those defaults gives about 385, which is then adjusted downward for small populations — that is why every row converges toward ~383.
The intuition: precision comes from the absolute number of people you hear from, not the fraction of the population they represent. A well-mixed sample of 400 tells you nearly as much about a million people as about ten thousand — the same way one spoonful tells you how a pot of soup tastes, regardless of the pot's size, as long as the soup is stirred.
How do you plan invites from a target sample size?
The table gives completed responses, not invitations. Divide by your expected response rate to size the send:
- 1Define the population (e.g., 4,000 active customers) and read the table → roughly 351–370 responses needed
- 2Estimate your response rate from past sends — internal employee surveys often see far higher rates than cold customer emails
- 3Divide: at an estimated 20% response rate, 370 needed ÷ 0.20 = 1,850 invitations
- 4If the required invites exceed your population, invite everyone and report the margin of error you actually achieved
- 5Send reminders rather than over-inviting — a second nudge typically recovers many non-responders
When your population is small — a 30-person team, a 150-customer niche product — skip the statistics and survey everyone. A census has no sampling error at all, and chasing '80 of 100' style targets matters only when you genuinely cannot reach the whole group.
What matters more than sample size?
Representativeness. Four hundred responses that all come from your happiest power users are worth less than 150 drawn evenly across your customer base. The math above assumes random sampling; in practice, who chooses to respond is the bigger threat to accuracy than how many respond. Watch for non-response bias (annoyed customers ignoring the email), channel bias (in-app prompts only reaching active users), and timing bias. Comparing the demographics of respondents against your known population is a quick sanity check.
One practical note on collection: because hitting 300–400 completed responses sometimes requires thousands of invites, response caps on free survey tools can bite. Formkii has no response limits on its free plan, so sample size targets are never constrained by the tool.
Frequently asked questions
How many responses do I need for a survey of 1,000 people?
At the standard 95% confidence level with a ±5% margin of error, you need about 278 completed responses from a population of 1,000. If you can tolerate a wider margin of error, you need fewer; for a tighter margin you need more.
Is 100 survey responses enough?
It depends on the population and the precision you need. For a population of 100, yes — 80 responses hits the standard ±5% bar. For a large population, 100 responses gives roughly a ±10% margin of error, which is fine for directional decisions but too loose for precise claims.
Why do large populations not need larger samples?
Statistical precision depends on the absolute number of responses, not the share of the population sampled. Once the population is large relative to the sample, adding more people to the population barely changes the math — which is why ~383 responses covers ±5% precision for 100,000 people or 100 million.
What is the difference between sample size and response rate?
Sample size is the number of completed responses you need; response rate is the share of invited people who actually complete the survey. To plan a send, divide the required sample size by your expected response rate — needing 370 responses at a 20% response rate means inviting about 1,850 people.