Skip to content
BestCalculators
Free online calculators and converters

Math & Statistics

Sample Size Calculator: How Many People to Survey

Work out the sample size a survey needs for a given confidence level and margin of error, corrected for the size of the population.

How often the true answer would fall inside your margin if the survey were repeated.

%

The ± band you can live with. Halving it roughly quadruples the sample needed.

%

Leave at 50 unless you have prior data. Fifty is the most cautious assumption, because it needs the largest sample.

Everyone the result should describe. Enter a very large number if the population is effectively unlimited.

%

To work out how many people you actually have to invite.

Responses needed

370people

Out of a population of 10,000. Without the finite-population correction the figure would be 384.

People to invite
1,234people

At a 30% response rate. This is the number that decides whether the survey is feasible.

Share of the population sampled
3.70%

Small populations need a much larger share than big ones — which is why national polls of 1,000 people work.

For a ±1% margin instead
4,900people

Precision is expensive: the sample grows with the square of the margin, so five times tighter costs twenty-five times more.

For a ±3% margin
965people
Margin you would get from 100 responses
9.80%

A useful reality check when the sample above is out of reach.

How to use this calculator

  1. Choose your Confidence level from 90 percent, 95 percent, or 99 percent based on how certain you need the results to be.
  2. Enter a Margin of error as a percentage, representing the plus-or-minus band you can live with in your final data.
  3. Leave the Expected proportion at 50 percent unless you have reliable prior data suggesting a different split.
  4. Input your Population size, using a very large number if the group you are studying is effectively unlimited.
  5. Enter your Expected response rate as a percentage to calculate how many survey invitations you actually need to send out.

Understanding Survey Sample Size

When planning any research project, determining the right sample size calculator metrics is the first operational hurdle. You cannot survey an entire nation, nor every customer who ever bought your product, so you must select a subset. That subset needs to be large enough to represent the whole without wasting time and budget on redundant interviews. The mathematical challenge lies in balancing statistical certainty against practical constraints.

A properly designed survey sample size rests on four statistical pillars: population size, variance, confidence level, and margin of error. When you adjust any one of these dials, the required number of respondents shifts dramatically. Ignoring even one of these variables introduces bias that can invalidate your entire data gathering effort, turning an expensive survey into an expensive guess.

The Math Behind the Numbers

At the core of standard statistical estimation is Cochran's formula, written as n₀ = z² × p(1−p) ÷ e². Here, z represents your chosen confidence score, p is the expected proportion of the attribute you are measuring, and e is your chosen margin of error. This baseline formula assumes an infinitely large population, which is why finite populations require an extra mathematical adjustment.

To correct for smaller groups, the calculation applies a finite population correction factor: n = n₀ ÷ (1 + (n₀ − 1) ÷ N), where N is your total population size. This adjustment acknowledges that sampling 500 people out of a total pool of 600 gives you far more comprehensive coverage than sampling 500 people out of six million. As your sample approaches the total population size, the required sample size drops rapidly because you are observing nearly everyone directly.

What the calculator is quietly doing behind the scenes is managing the extreme caution of 50 percent proportions. When you leave the expected proportion at 50 percent, the product of p(1−p) reaches its absolute maximum of 0.25. This forces the formula to assume the highest possible variance in answers, ensuring your statistically significant sample size remains robust even if your audience splits down the middle on every question you ask.

Margin of Error and Confidence Intervals

Your margin of error calculator results define the boundaries of your findings. If 60 percent of your respondents say they prefer your new logo, and your margin is 3 percent, you can be reasonably sure the true population preference lies between 57 percent and 63 percent. The confidence level dictates how often that guarantee holds true if you repeated the survey under identical conditions hundred of times.

Confidence LevelZ-ScoreTypical Use Case
90%1.645Internal employee feedback and rough pulse checks
95%1.960Standard market research and academic polling
99%2.576Clinical trials, high-stakes finance, and safety compliance

Choosing a 99 percent confidence level means accepting a wider interval or needing a massive number of completions to compensate. Most commercial projects settle on 95 percent as the sweet spot between financial cost and statistical rigor. When calculating your confidence interval calculator requirements, always weigh the cost of being wrong against the cost of collecting more data.

Invitation Rates and Response Realities

Knowing how many completed surveys you need is only half the battle. People ignore emails, hang up on phone calls, and abandon web forms halfway through. To bridge the gap between completed responses and people to invite, the system divides your target sample by your expected response rate. If you need 400 completions and anticipate a 20 percent response rate, you must send 2,000 invitations.

A common mistake in survey design is assuming a high response rate without evidence. Researchers often plug in an optimistic 50 percent response rate for cold email outreach, only to achieve 4 percent in reality. This leaves them drastically underpowered, with a tiny fraction of the data required to maintain their intended margin of error. Always use historical benchmarks from past outreach efforts when estimating your response rates.

When your response rate falls short, your effective sample size shrinks, which blows out your margin of error. If you planned for 400 responses at plus-or-minus 5 percent but only collect 100 responses, your actual margin of error expands significantly. You can reverse-engineer this degradation by examining what level of precision 100 responses actually buys you under standard variance assumptions.

Limitations and When to Seek Expert Help

These mathematical models assume random sampling where every member of the target population has an equal chance of selection. If your sampling method suffers from selection bias—such as posting a poll on a public social media page where only angry or highly enthusiastic customers reply—no sample size formula will save the validity of your data. Large, biased samples are simply large, biased errors.

You should not rely on standard formulas when dealing with complex stratified sampling, cluster sampling, or multi-tiered organizational hierarchies. In those scenarios, standard random formulas will understate your true error margins. When your research carries legal, regulatory, or multi-million-dollar financial consequences, bypass standard automated estimators and consult a professional statistician or data scientist to design your sampling frame.

The formula

Cochran: n₀ = z² × p(1−p) ÷ e²finite population correction: n = n₀ ÷ (1 + (n₀ − 1) ÷ N)people to invite = n ÷ response ratemargin from a sample: e = z × √(p(1−p) ÷ n)

Frequently asked questions

What is the difference between confidence level and margin of error?

Confidence level measures the certainty with which your results reflect the true population, usually set at 95 percent. Margin of error defines the acceptable percentage boundary around your results, showing how much your actual figures might vary. Together, they frame the reliability of your survey data.

Why does the calculator default to a 50 percent expected proportion?

Setting the expected proportion to 50 percent maximizes statistical variance in Cochran's formula. This is the most cautious assumption you can make because a 50-50 split requires the largest possible sample size to achieve your desired precision. If you have solid historical data showing a 90-10 split, you can safely lower this number to reduce your required sample.

How does the population size affect my required sample?

For very large populations like entire countries or global customer bases, increasing the population total stops changing the required sample size. However, for smaller finite groups like employees in a mid-sized company, the finite population correction significantly lowers the number of responses you need. Once your sample size approaches your total population, you are effectively conducting a census rather than a sample.

Why do I need to send out more invitations than my target sample size?

No survey achieves a 100 percent completion rate due to ignored messages, opt-outs, and abandoned forms. Accounting for your expected response rate ensures you invite enough people to hit your target number of completed responses. Failing to factor in response rates leaves your project underpowered with too little data.

Can my sample size be too large?

Technically, a larger sample size always reduces your margin of error and increases precision. Practically speaking, however, collecting data beyond a certain point yields diminishing statistical returns at a very high financial cost. Once your sample size satisfies your required confidence and error limits, additional responses waste time and budget.

Last reviewed . Results are for general guidance and are not professional advice.