What Is Sample Size?
Sample size is the number of survey responses you need for your results to be statistically reliable — that is, for your findings to reflect what your entire audience actually thinks, not just the few people who happened to answer.
Ask 5 colleagues and you’ll get 5 opinions. Ask a properly calculated sample of your 10,000 customers, and you can state things like “62% of customers are satisfied, ±5%” with known confidence. That’s the difference between an anecdote and a finding.
How the Sample Size Calculator Works
The calculator uses the standard two-step approach:
Step 1 — base sample size (for a large population):
n₀ = z² × p × (1 − p) ÷ e²
Step 2 — adjust for a finite population (when you know your population size N):
n = n₀ ÷ (1 + (n₀ − 1) ÷ N)
Where:
- z = z-score for your confidence level (1.96 at 95%)
- p = expected proportion (0.5 is the conservative default)
- e = margin of error as a decimal (0.05 = ±5%)
The result is rounded up to the next whole response.
Worked example. Population 1,000 customers, 95% confidence, ±5% margin of error, 50% expected proportion:
- n₀ = (1.96² × 0.5 × 0.5) ÷ 0.05² = 384.16
- n = 384.16 ÷ (1 + 383.16 ÷ 1,000) = 277.8 → 278 responses
How to use the calculator:
- Enter your population size — the total number of people your survey represents (all active customers, all employees, all members). Don’t know it, or is it huge? Leave it blank — the calculator treats it as an unknown/large population and skips the finite-population adjustment.
- Choose your confidence level — 95% is the standard for most surveys.
- Set your margin of error — ±5% is the common default.
- Leave expected proportion at 50% unless you have a solid estimate (see below).
- Read your result: the minimum number of complete responses needed, plus what that represents as a percentage of your population.
The calculator validates input — the margin of error must be greater than 0%, and a population size, if entered, must be at least 1.
Choosing Your Inputs
Confidence level
How sure you want to be that your result falls within the margin of error:
| Confidence level | Z-score | Responses needed (unknown population, ±5%) | Typical use |
|---|---|---|---|
| 90% | 1.645 | 271 | Internal, exploratory, quick reads |
| 95% | 1.96 | 385 | The standard for most surveys |
| 99% | 2.576 | 664 | High-stakes, published, or regulated research |
Margin of error
How much wiggle room you’ll accept around your results. Smaller margins cost exponentially more responses:
| Margin of error | Responses needed (unknown population, 95%) | Reading it in practice |
|---|---|---|
| ±10% | 97 | Rough directional signal only |
| ±5% | 385 | If 60% say yes, the true figure is likely 55–65% |
| ±3% | 1,068 | Tighter claims for decisions with real money attached |
| ±1% | 9,604 | Near-census precision — rarely worth it for surveys |
Expected proportion
How split you expect the answers to be. 50% is the safest choice because it assumes maximum disagreement — if you don’t know, leave it there. But if you’re re-measuring something you’ve surveyed before (say, last wave showed 80% satisfaction), entering 80% legitimately reduces the required sample: the same 95% / ±5% setup then needs 246 responses instead of 385. The further from an even split, the fewer responses you need.
Population size (optional)
Here’s the counterintuitive part — sample size barely grows with population:
| Population | Responses needed (95%, ±5%) | Share of population |
|---|---|---|
| 100 | 80 | 80.0% |
| 1,000 | 278 | 27.8% |
| 10,000 | 370 | 3.7% |
| 100,000 | 383 | 0.4% |
| Unknown / very large | 385 | — |
Past a few thousand people, the requirement flattens out around 385 — which is why a national poll of 330 million people works with ~1,000 respondents. The flip side: small populations need to answer a much larger share. Surveying a 100-person team still takes 80 of them for ±5% precision — at that point, consider just surveying everyone.
From Target to Fieldwork
Your sample size is a quota of complete responses. To hit it:
- Convert the target into invitations. Most people invited won’t respond. If you need 278 responses and expect a 20% response rate, you’ll need to reach ~1,400 people. Work that math in reverse with the survey response rate calculator.
- Over-sample slightly. Plan for inattentive or duplicate responses; a small buffer above your target protects the validity of the final dataset.
- Design for completion. Every abandoned response is a wasted invitation. One-question-at-a-time surveys consistently finish at higher rates than long scrolling forms — this is exactly the flow Formware is built for.
- Distribute where your audience is. Share by email, link, or QR code, or send the survey through WhatsApp if that’s where your audience lives.
- Track progress against your quota. Watch cumulative form responses as they come in, and tag each distribution channel with UTM parameters so you know which channel is actually delivering your respondents.
- Mind your subgroups. If you plan to report results per segment — per region, per plan, per team — each subgroup needs its own valid sample size. Comparing 90 responses from Region A against 30 from Region B isn’t a valid comparison; multiply your target by the number of segments that need to stand on their own.
When to Use a Sample Size Calculator
- Customer research — satisfaction studies, concept tests, brand tracking before committing to findings
- NPS and feedback programs — knowing your quota helps you judge whether a wave’s results are trustworthy yet (for scoring them, see the NPS calculator)
- Employee research — engagement and pulse surveys across a known headcount
- Market sizing and validation — before pricing decisions, feature bets, or launches
- Community, academic, and public-sector research — anywhere a finding needs to survive scrutiny
Start Collecting Responses That Count
The calculator tells you how many responses you need — Formware is how you get them. The market research use case guide covers planning end to end, or browse the research survey templates and copy one straight into your account: edit the questions, publish, and start working toward your quota.
Frequently Asked Questions
How is sample size calculated?
With a two-step formula: first the base size n₀ = z² × p(1−p) ÷ e², then a finite-population correction n = n₀ ÷ (1 + (n₀−1) ÷ N) when you know your population. Example: 1,000 people at 95% confidence and ±5% error → 278 responses.
Is this sample size calculator free?
Yes — completely free, no signup required. If you also need to collect the responses, the survey template below is free to copy into a Formware account.
What if I don't know my population size?
Leave it blank — the calculator treats it as an unknown/large population and uses the base formula only. That result (385 at 95% / ±5%) is the maximum you’d ever need for any population at those settings, so it’s a safe ceiling.
Why does sample size barely grow with population size?
Statistical sampling scales with variability, not headcount. The finite-population correction shrinks the requirement as the population gets smaller relative to the sample — which is why 10,000 people need 370 responses but 100 people need 80. Past a few thousand, the number plateaus around 385 at 95% / ±5%.
What margin of error should I use?
±5% is the standard for most surveys; ±3% when decisions carry real stakes (pricing, positioning, published research); ±10% only for quick directional reads. Remember what it means: at ±5% with 60% agreeing, the true figure likely falls between 55% and 65%.
What's the expected proportion, and should I change it from 50%?
It’s how split you expect the answers to be. 50% is the conservative default because it assumes maximum disagreement. If you have a solid estimate from earlier research — say 80% satisfied — entering it reduces the required sample (246 instead of 385 at 95% / ±5%). When in doubt, leave it at 50%.
What's the difference between confidence level and margin of error?
Confidence level is how sure you want to be; margin of error is how precise you want to be. 95% / ±5% means: if you repeated the survey many times, 95% of the repeats would land within ±5 points of the true population value. Raising either one increases the required sample.
How many people do I need to invite to hit my target?
Divide your sample size by your expected response rate. Need 278 responses at a 20% response rate → invite roughly 1,400 people. The survey response rate calculator works through this with your actual numbers.
Can I use this for subgroups?
Yes, but calculate each subgroup separately. If you want to report valid results for three regions independently, you need the target sample size per region — roughly triple the total. A single overall sample of 300 doesn’t give you three valid samples of 100.
Does this work for NPS, CSAT, or CES surveys?
For planning response volumes, yes — the calculator tells you how many responses you need for the proportions in your results to be reliable. Those metrics are themselves built from proportions, so a properly sized sample underpins them. Use the NPS, CSAT, and CES calculators to score the results once they’re in.
