Sample Size Calculator
How many people do you need to survey before you can trust the result? Too few and the margin of error swallows your finding; too many and you waste time and money. The answer comes from Cochran's formula, which balances three things: how confident you want to be, how precise the result must be, and how varied the population is.
Pick a confidence level and a margin of error, enter your best guess at the proportion you are measuring (or leave it at 0.5 for the safest answer), and add the population size if it is small. The calculator tells you how many respondents you need — always rounded up, because 384.16 people is not an option.
Respondents needed
How It Works
The calculation is Cochran's formula:
n0 = z² × p(1 − p) / e²
Here z is the critical value for your confidence level (1.645 for 90%, 1.96 for 95%, 2.576 for 99%), p is the population proportion you are measuring, and e is the margin of error as a decimal (5% is 0.05). The answer is always rounded up: a formula result of 384.16 means you need 385 people, because 384 would leave you just short of the margin you asked for.
The proportion p matters because p(1 − p) measures how varied the population is. It is largest when p = 0.5 (0.25) and shrinks towards 0 as p approaches 0 or 1 — a population where almost everyone agrees needs a smaller sample than a 50/50 split. That is why 0.5 is the safe default: it gives the largest sample size, so you are covered whatever the true proportion turns out to be.
When the population is finite and your sample would be a noticeable share of it, the finite population correction applies:
n = n0 / (1 + (n0 − 1) / N)
It shrinks the required sample, because surveying a large fraction of a small population leaves less unknown. The correction is applied to the unrounded n0 and the final answer is rounded up once at the end. If the population is very large — or unknown — the correction makes almost no difference and is skipped.
One pattern is worth knowing: halving the margin of error needs four times the sample, because e is squared in the formula. Going from ±5% to ±2.5% at 95% confidence takes you from 385 respondents to 1,537.
Worked Example
You want a survey accurate to ±5% at 95% confidence, with no idea of the true proportion, so p = 0.5. Cochran's formula gives n0 = 1.96² × 0.5 × 0.5 / 0.05² = 3.8416 × 0.25 / 0.0025 = 0.9604 / 0.0025 = 384.16, rounded up to 385 respondents.
If the population is 10,000 people, the finite population correction gives n = 384.16 / (1 + 383.16 / 10,000) = 384.16 / 1.038316 = 369.98, rounded up to 370 respondents — a saving of 15 people. And if you need much more precision — 99% confidence with a ±3% margin — the formula gives n0 = 2.576² × 0.25 / 0.03² = 6.635776 × 0.25 / 0.0009 = 1,843.27, rounded up to 1,844 respondents.
Assumptions
- Simple random sampling. Every member of the population is equally likely to be chosen. Stratified, clustered or convenience sampling needs a larger sample than this formula gives.
- Normal approximation. Cochran's formula uses the normal (z) distribution. It is the standard approach for survey sample sizes and is accurate for the sample sizes it produces.
- p = 0.5 is the conservative choice. It maximises p(1 − p) and therefore the sample size. If you genuinely know the proportion is far from 0.5, entering it gives a smaller — but still valid — sample.
- The correction uses the unrounded n0. The finite population correction is applied before rounding, and the final answer is rounded up once. Rounding twice would overstate the sample by one in cases like the worked example.
- Non-response is not modelled. The answer is the number of completed responses you need. If only half the people you invite reply, invite twice as many.
- Margin of error must be under 100% and the proportion strictly between 0 and 1 — a proportion of exactly 0 or 1 would mean there is nothing to measure.
Sources: No external data is used on this page — every figure is calculated from the values you enter, using Cochran's formula (n₀ = z² · p(1 − p) / e²) with z = 1.645 (90%), 1.96 (95%) or 2.576 (99%), and the standard finite population correction. Last reviewed: 15 November 2026.
Built and tested by SumKitTool — see how we check our numbers.
How to Use This Calculator
Pick the confidence level and enter the margin of error you can live with. Leave the proportion at 0.5 unless you have a good estimate — 0.5 gives the largest, safest sample. Add the population size only if it is small enough that your sample would be a noticeable share of it. The headline result is the number of respondents you need; the table shows the unrounded formula result and the correction, if one was applied.
Try the worked example: 95% confidence, ±5%, p = 0.5 gives 385. Add a population of 10,000 and it drops to 370. Then try ±3% at 99% to see it jump to 1,844 — precision is expensive.
Tips and Practical Advice
Precision costs far more than confidence. Moving from 95% to 99% confidence at ±5% takes the sample from 385 to 664, but halving the margin to ±2.5% at 95% takes it to 1,537. If the budget is tight, loosen the margin before you loosen the confidence.
Always plan for non-response. A survey that needs 385 completed responses and gets a 40% response rate needs about 963 invitations — and the people who do not respond are often systematically different from those who do, which no sample-size formula can fix.
The finite population correction only matters when the sample is a real share of the population. For a population of 10,000 it saves 15 people out of 385; for a population of 1,000,000 it saves less than one. If your population is large, leave the field blank.
Frequently Asked Questions
Why is 0.5 the default population proportion?
Because it gives the largest possible sample size, which is the safe answer when you do not know the true proportion. The term p(1 − p) in Cochran's formula measures how varied the population is, and it peaks at p = 0.5 (0.25). If the true proportion is 0.1 or 0.9, p(1 − p) is only 0.09, so a smaller sample would do — but using 0.5 guarantees you are covered either way.
What is the finite population correction?
It is an adjustment for when your sample would be a noticeable share of a small, known population. Surveying 385 people out of 10,000 leaves less unknown than surveying 385 out of 10 million, so the correction shrinks the required sample — from 385 to 370 in the 10,000-person example. The formula is n = n₀ / (1 + (n₀ − 1) / N). For very large populations the correction is negligible and can be skipped.
Why is the answer always rounded up?
Because you cannot survey a fraction of a person, and rounding down would leave you short of the margin of error you asked for. A formula result of 384.16 means 384 people is not quite enough — you need 385. Rounding up is the only direction that keeps the promised precision.
What if my population is small?
Enter it in the population size field and the finite population correction does the work. At 95% confidence with a ±5% margin, a population of 1,000 needs 278 respondents instead of 385, and a population of 100 needs only 80 — because surveying 80 out of 100 people leaves very little unknown. Without the correction you would waste effort over-sampling a small group.
Last updated: