How many people do you need to survey? Use our free sample size calculator below, then learn how the Cochran formula works, when to apply the finite population correction, and how to turn a sample size into the number of survey invitations you actually need to send.
Sample Size Calculator
Based on the Cochran formula, with the finite population correction applied when a population size is entered.
Table of Contents
- What Is Sample Size?
- The Cochran Formula for Sample Size
- Adjusting for a Small Population (Finite Population Correction)
- How to Calculate Sample Size Step by Step
- Worked Examples
- Sample Size Tables (Quick Reference)
- How to Choose Your Inputs
- Sample Size vs. Number of Invitations
- Common Mistakes to Avoid
- Conclusion
- FAQ
What Is Sample Size?
The sample size is the number of completed responses you need so that your survey results reflect the whole population within an acceptable level of precision. Survey everyone and you get the exact answer, but that is rarely possible or affordable. Survey too few people and random chance can push your results far from the truth.
The right sample size depends on three things:
- Confidence level: how sure you want to be that the true value falls inside your range (usually 95%).
- Margin of error: how far your result may be from the true value (for example ±5 percentage points). See our margin of error calculator for a deep dive.
- Variability: how split the answers are expected to be. A 50/50 split needs the largest sample.
Example: to say “62% of our customers would recommend us, ±5% at 95% confidence”, you need roughly 385 completed responses from a large customer base.
The Cochran Formula for Sample Size
The most widely used sample size formula for surveys was published by statistician William G. Cochran. It estimates the sample needed to measure a proportion (a percentage of people who choose an answer) in a large population:
n0 = Z² × p × (1 − p) / e²
Where:
- n0 = required sample size for a large (or unknown) population.
- Z = Z-score for your confidence level (1.96 for 95%).
- p = expected proportion, as a decimal. Use 0.5 when you don’t know: it gives the most conservative (largest) sample.
- e = desired margin of error, as a decimal (0.05 for ±5%).
Z-Scores for Common Confidence Levels
| Confidence Level | Z-Score | Z² |
|---|---|---|
| 80% | 1.282 | 1.642 |
| 90% | 1.645 | 2.706 |
| 95% | 1.96 | 3.842 |
| 99% | 2.576 | 6.635 |
Why p = 0.5 Is the Safe Default
The term p × (1 − p) is largest when p = 0.5 (it equals 0.25). Any other value gives a smaller product and therefore a smaller sample. If a previous survey tells you the answer is likely around 70%, you can use p = 0.7 (product = 0.21) and save some responses. If you have no prior data, stick with 0.5.
Adjusting for a Small Population (Finite Population Correction)
The Cochran formula assumes the population is very large. When your population is small, say 800 employees or 2,000 customers, you need fewer responses. Apply the finite population correction (FPC):
n = n0 / (1 + (n0 − 1) / N)
Where N is the total population size. As a rule of thumb, the correction matters when your sample would be more than 5% of the population. For populations over 100,000 the result barely changes, which is why national polls of millions of people still use around 1,000 respondents.
How to Calculate Sample Size Step by Step
- Pick a confidence level and look up its Z-score (95% → 1.96).
- Choose a margin of error and convert it to a decimal (±5% → 0.05).
- Estimate the proportion p (use 0.5 if unsure).
- Compute n0 with the Cochran formula.
- If you know the population size N, apply the finite population correction.
- Round up to the next whole number. Always round up, never down.
- Divide by your expected response rate to find how many people to invite.
Worked Examples
Example 1: Large Population, 95% Confidence, ±5%
- Z = 1.96, p = 0.5, e = 0.05
- n0 = 1.96² × 0.5 × 0.5 / 0.05² = 3.8416 × 0.25 / 0.0025 = 384.16
- Rounded up: 385 responses
This is the famous “385” you see in many survey guides. It applies to any large population, whether it’s 50,000 or 50 million people.
Example 2: Known Population (Customer Survey)
A Shopify store wants to survey its 2,000 customers about satisfaction at 95% confidence and ±5%.
- n0 = 384.16 (from Example 1)
- n = 384.16 / (1 + 383.16 / 2,000) = 384.16 / 1.1916 = 322.4
- Rounded up: 323 responses
If the store expects a 20% response rate on a post-purchase survey, it should invite about 323 / 0.20 = 1,615 customers.
Example 3: Using a Prior Estimate
A previous survey found that 70% of people would consider online-only banking. You want to check this in your community with 90% confidence (Z = 1.645) and a ±5% margin of error.
- n0 = 1.645² × 0.7 × 0.3 / 0.05² = 2.706 × 0.21 / 0.0025 = 227.3
- Rounded up: 228 responses
With p = 0.5 you would have needed 271, so the prior estimate saves about 16% of the responses.
Sample Size Tables (Quick Reference)
Large Population (p = 0.5)
| Margin of Error | 90% Confidence | 95% Confidence | 99% Confidence |
|---|---|---|---|
| ±10% | 68 | 97 | 166 |
| ±5% | 271 | 385 | 664 |
| ±3% | 752 | 1,068 | 1,843 |
| ±2% | 1,692 | 2,401 | 4,147 |
| ±1% | 6,765 | 9,604 | 16,587 |
By Population Size (95% Confidence, ±5%, p = 0.5)
| Population Size (N) | Required Sample Size |
|---|---|
| 100 | 80 |
| 500 | 218 |
| 1,000 | 278 |
| 5,000 | 357 |
| 10,000 | 370 |
| 100,000 | 383 |
| 1,000,000+ | 385 |
Notice how quickly the curve flattens: going from 10,000 to 1,000,000 people adds only 15 responses. Precision depends on the size of the sample, not on the share of the population you survey.
How to Choose Your Inputs
Confidence Level
- 95% is the standard for most business and academic surveys.
- 90% is fine for quick internal decisions, product feedback or exploratory research.
- 99% is for high-stakes decisions (medical, legal, financial) where being wrong is costly.
Margin of Error
- ±5% is a common default for customer and market surveys.
- ±3% or less if you need to detect small differences, such as comparing two close options.
- ±10% can be acceptable for early-stage directional feedback.
Halving the margin of error quadruples the sample size, because e is squared in the formula. Going from ±5% to ±2.5% takes you from 385 to 1,537 responses.
Subgroups
If you plan to analyze segments separately (for example new vs. returning customers, or each country), each segment needs its own sample size. Three segments at ±5% means roughly 3 × 385 responses, not 385 in total.
Sample Size vs. Number of Invitations
The calculator gives you the number of completed responses. To find out how many people to invite:
Invitations = Sample size / Response rate
Typical response rates vary a lot: 5–15% for cold email surveys, 20–40% for engaged customers, and much higher for on-site or in-app forms shown at the right moment. Raising your response rate is often cheaper than buying more reach. Our guide on how to increase survey response rate covers proven tactics, and conversational one-question-at-a-time forms built with Veeform typically get more people to the end.
Common Mistakes to Avoid
- Confusing invitations with responses: 385 invitations will not give you 385 responses.
- Ignoring non-response bias: a big sample does not fix a biased one. If only your happiest customers answer, results will be skewed whatever the size.
- Using convenience samples: the formula assumes random sampling. Surveying only your social media followers does not represent all customers.
- Forgetting subgroups: plan sample size for the smallest group you want to report on.
- Rounding down: always round up to stay within your target margin of error.
- Using the proportion formula for averages: if you measure a mean (like average spend), use n = (Z × σ / e)², where σ is the standard deviation and e the margin of error in the same unit.
Conclusion
A good sample size balances precision and cost. With the Cochran formula you can work out exactly how many responses you need:
- Start with 95% confidence, ±5% margin of error and p = 0.5: 385 responses for any large population.
- Apply the finite population correction when surveying a small, known group.
- Divide by your response rate to plan invitations.
Once you have your target, build a survey people actually want to finish. Create your survey with Veeform and check your final precision with our margin of error calculator.
FAQ
1. What is the Cochran formula?
The Cochran formula, n0 = Z² × p × (1 − p) / e², calculates the sample size needed to estimate a proportion in a large population for a given confidence level (Z), expected proportion (p) and margin of error (e).
2. What sample size do I need for 95% confidence and a 5% margin of error?
385 completed responses for a large or unknown population, assuming p = 0.5. For smaller populations it’s lower: 278 for 1,000 people and 218 for 500.
3. Why use 0.5 as the proportion?
0.5 maximizes p × (1 − p), so it gives the largest, most conservative sample size. Use it whenever you have no reliable prior estimate.
4. When should I use the finite population correction?
When your sample would represent more than about 5% of the total population. The correction n = n0 / (1 + (n0 − 1) / N) reduces the required sample.
5. Does a bigger population need a much bigger sample?
No. Beyond roughly 20,000 people the required sample barely changes. At 95% confidence and ±5%, both 100,000 and 10 million people need about 385 responses.
6. What is the difference between sample size and sampling error?
Sample size is the number of responses you collect. Sampling error (or margin of error) is the uncertainty that comes from surveying a sample instead of everyone. A larger sample means a smaller sampling error.
7. How do I calculate sample size for a mean instead of a percentage?
Use n = (Z × σ / e)², where σ is the estimated standard deviation and e is the acceptable error in the same unit (for example ±$5).
8. How many people should I invite to my survey?
Divide your required sample size by your expected response rate. For 385 responses at a 25% response rate, invite about 1,540 people.
9. Is a sample of 100 enough?
For a large population, 100 responses give a margin of error of about ±9.8% at 95% confidence. That can be enough for directional feedback, but not for precise comparisons.
10. Can I use this calculator for NPS or rating scale surveys?
Yes, for estimating the share of people giving a given answer (such as the percentage of promoters or of people choosing 4 or 5 on a 1-5 scale). For the NPS score itself, which combines two proportions, plan for a somewhat larger sample. You can then compute your score with our NPS calculator.
