---
title: "Margin of error calculator for surveys — free"
description: "Calculate a survey's margin of error from the number of responses, confidence level and population size. Explanation, examples, how to read it."
lang: en
url: https://fidestat.com/en/tools/margin-of-error-calculator/
---

# Margin of error calculator

Check how precise your survey results are. Enter the number of responses and the confidence level — the calculator shows how many percentage points the result may be off from the true value.

- [Calculate now](https://fidestat.com/en/tools/margin-of-error-calculator/#alat)

The margin of error shows how far a survey result may be from the true value in the whole population simply because you didn't ask everyone. The calculator above works it out from the number of responses and the confidence level, and optionally from the population size.

## How to read the result

If 62% of respondents are satisfied and the margin is ±4.9 points, at 95% confidence the true share of satisfied people is between 57.1% and 66.9%. Always report three numbers together: the result, the margin and the number of responses. “62% (±4.9 points, n = 400)” tells the reader at once how much to trust the figure.

## How it is calculated

- e = z · √(p(1 − p) / n), where n is the number of responses, z the value for the confidence level (1.645 for 90%, 1.960 for 95%, 2.576 for 99%), and p = 0.5.
- If a population N is entered, the margin is multiplied by the correction √((N − n) / (N − 1)). When you survey a large share of a small population, this correction reduces the margin considerably.

The value p = 0.5 gives the largest possible margin, so the margin shown is a safe upper bound for any question with two answers (yes/no, satisfied/not satisfied).

## Margin for common response counts

| Responses | 90% confidence | 95% confidence | 99% confidence |
| --- | --- | --- | --- |
| 100 | ±8.2 | ±9.8 | ±12.9 |
| 200 | ±5.8 | ±6.9 | ±9.1 |
| 400 | ±4.1 | ±4.9 | ±6.4 |
| 1,000 | ±2.6 | ±3.1 | ±4.1 |
| 2,000 | ±1.8 | ±2.2 | ±2.9 |

Values are in percentage points, for a large population. Going from 100 to 400 responses halves the margin; halving it again takes 1,600 responses.

## What the margin does not measure

The margin of error describes only **random** sampling error. It does not cover:

- **selection bias** — if only app users were invited, the result says nothing about people who don't use the app;
- **non-response** — if mostly dissatisfied (or only delighted) people answered, the result is skewed;
- **question wording** — a leading question produces a wrong result with a very small margin;
- **order and fatigue** — answers at the end of a long survey are less careful.

A small margin is necessary, but not sufficient. Well-written questions and a clearly defined group of respondents matter just as much.

## When you need a smaller margin

If you use the result for a decision with a threshold (“we launch delivery if more than half want it”) and the result is close to that threshold, the margin must be smaller than the gap between the result and the threshold. Otherwise the survey gives no answer — and that is useful to know: you need more responses, not a new assumption. The sample size calculator shows how many responses you need for the margin you want.

## Examples

- A supermarket received 400 responses and 58% of customers say they would use delivery. The margin is ±4.9 points, so the true share is likely between 53% and 63% — a majority, but not a convincing one.
- HR survey: 180 of 250 employees responded. With a population of 250 the margin is about ±3.9 points, instead of ±7.3 for a large population.
- A restaurant has 120 responses and 34% rate it “excellent”. The margin is ±8.9 points: the result gives a rough picture, but is not precise enough to compare with last month.

## Common mistakes

- Comparing two results only by their margins. If the ranges partly overlap, the difference can still be significant — use the two-proportion z-test for that.
- Reporting the whole-sample margin for subgroups (for example, only young respondents).
- Reading ±5% as “5% of the result”. For a 40% result it means 35–45%, not 38–42%.
- Presenting the margin as the only quality measure of a survey, without saying who was invited and how many people answered.

## Frequently asked questions

### What does a “±5% margin of error” mean?

If a survey shows 40%, the true value in the population is, at the chosen confidence level, likely between 35% and 45%. The margin is expressed in percentage points, not as a percentage of the result.

### What is the margin of error for 100, 400 and 1,000 responses?

At 95% confidence and a large population: about ±9.8 points for 100 responses, ±4.9 for 400 and ±3.1 for 1,000.

### Why is the margin the same whatever the survey result?

The calculator uses p = 0.5, which gives the largest possible margin. For results close to 0% or 100% the true margin is somewhat smaller, so the value shown is a safe upper bound.

### Does the margin apply to questions not everyone answered?

No. For each question, use the number of people who actually answered it. If only 150 of 400 respondents saw a question, the margin for that question is about ±8 points.

### Does the margin of error cover badly worded questions?

No. It measures only the random error from asking some people instead of everyone. Leading questions, the wrong respondents or a low response rate create errors the margin does not show.

### When should I enter the population size?

When you surveyed a large share of a small group — for example 200 of 500 employees. The true margin is then smaller than for a large population, and the calculator takes that into account.

- [Start for free](https://app.fidestat.com/registracija)
