Margin of Error Calculator

Find out how much your survey result can move once you read it back onto the whole population. Enter your population size, the responses you collected, and a confidence level, and the band updates as you type.

Free to use
No sign-up
Runs in your browser

Calculate your margin of error

Your inputs

Everyone the result is meant to speak for, not only the people you contacted.

How often the true answer would land inside the band if you ran the survey again. 95% is the usual choice.

Completed responses you actually collected. Partial and abandoned answers do not count.

Nothing you type here leaves your device.

Margin of error

Enter a population size and a sample size to see how accurate your result is.

Key terms

Population size

Total number of people in the population you are researching. If you survey your organization, population size is the total number of employees.

Confidence level

A percentage that indicates how sure you can be that a certain amount of the population will select an answer. A 95% confidence level, for example, means that the outcomes between x and y are 95% certain.

Sample size

A value that tells how many people you need to interview to get results that will represent the target population as accurately as possible.

This band covers sampling error only, and assumes a simple random sample with a 50/50 population proportion. Question wording, who chose to answer, and who stayed silent all sit outside it.

How accurate is your survey result?

Surveying is a balancing act: you question a smaller group and use their answers to describe a much larger one. The margin of error puts a number on what that shortcut costs you. A 60% "yes" with a 5% margin means 55% to 65% of the wider population would agree, so a rival option sitting at 58% is not behind at all.

1
Count what you collected
Take your completed responses and the population they came from. Both go into the box above.
2
Read the band
Add and subtract the result from any percentage in your findings to see how far it could really sit.
3
Report it alongside
Quote the margin next to your headline number. Without it, nobody reading your report can tell how firm that number is.

The margin of error formula

Nothing here is invented for this page. It is the textbook confidence interval for a proportion, so you can check any number the calculator gives you.

Margin of Error Formula
  1. z is the z-score for your confidence level: 1.28 at 80%, 1.65 at 90%, 1.96 at 95%, 2.58 at 99%.
  2. p is the population proportion, held at 0.5 here because a 50/50 split produces the widest band and never flatters your result.
  3. n is your sample size, and the square root under it is why quadrupling your responses only halves the margin.
  4. The finite population correction then narrows the band using your population size N, which matters when you have surveyed a large share of a small group.

Free, private, and built for research

Most calculators of this kind want your email before they will show a number, and many run the arithmetic on someone else's server. This one does neither.

  • 01

    Runs in your browser

    No API call sits behind this page. Your response counts and study parameters stay on your machine.

  • 02

    No account, no email wall

    Change an input, read the number, close the tab. Nothing is logged and nothing is asked of you.

  • 03

    Working you can check

    The formula is printed above and the same inputs always return the same figure, so you can defend it in a method section.

Who reaches for a margin of error calculator

Anyone who has to say out loud how much a result can be trusted.

01
Researchers
Report the confidence interval next to every headline figure, the way a reviewer expects to see it.
02
Product teams
Check whether two options really differ before a preference survey decides your roadmap.
03
Marketing and CX
Know whether this quarter's score genuinely moved or simply wobbled inside the band.
04
HR and people teams
Say how far an engagement result could sit from the mood of the whole company.
05
Students and academics
Produce the confidence interval your write-up has to state and justify.
06
Agencies
Show a client what their sample actually bought them in precision, before they read too much into it.

Frequently asked questions

What is a margin of error?

It is the band around your survey result, written in percentage points. If 60% of respondents said yes and your margin of error is 5%, the true figure for the whole population sits somewhere between 55% and 65%. The narrower the band, the more precisely your sample speaks for everyone you did not survey.

How is the margin of error calculated?

It divides the z-score for your confidence level, multiplied by the square root of p times (1 minus p), by the square root of your sample size, then applies the finite population correction so a sample drawn from a small population is not penalised. The population proportion p is held at 0.5, the value that produces the widest and therefore most cautious band. The formula image on this page shows the full working.

What is a good margin of error for a survey?

5% at a 95% confidence level is the working standard for most survey research. Below 3% you are into the territory of published polling and national studies, and the sample sizes get expensive fast. Above 10% the result is directional at best: fine for a pulse check, weak ground for a decision that costs money.

How do I reduce my margin of error?

Collect more responses. The relationship is a square root, so the returns shrink as you go: quadrupling your sample halves the margin of error. Accepting a lower confidence level narrows the band too, though you trade certainty for it. Where the population is small, sampling a larger share of it also helps through the finite population correction.

Why does my margin of error read 0%?

Because you have surveyed the entire population. When sample size equals population size there is no sampling error left to measure: you have a census rather than a sample, and the result speaks for the group exactly.

Does a small margin of error mean my survey is accurate?

Only in one specific sense. It measures sampling error, the wobble that comes from questioning some people rather than all of them. It says nothing about leading questions, respondents who differ from the people who ignored you, or answers given to look good. A tight margin of error on a biased sample is a precise measurement of the wrong thing.

Is anything I type stored?

No. The calculation is arithmetic that runs in your browser. There is no API call behind it, no account, and no record of what you entered.
Scripts are blocked. This site won’t work properly. If you’re using Brave, click the Shields icon and turn off Block scripts. Otherwise disable your ad blocker for this site.