---
title: "Sample size calculator — how many survey responses you need"
description: "Free sample size calculator for surveys: confidence level, margin of error and population size. Formula, worked examples and common mistakes."
lang: en
url: https://fidestat.com/en/tools/sample-size-calculator/
---

# Sample size calculator

Find out how many respondents you need for survey results that hold for the whole population — at your chosen confidence level and margin of error. Free, no sign-up, formula explained.

- [Calculate now](https://fidestat.com/en/tools/sample-size-calculator/#alat)

Sample size is the number of respondents who need to answer for survey results to hold, with known precision, for the whole group you are studying. The calculator above works it out from three inputs: how sure you want to be (confidence level), how far the result may be off (margin of error) and, optionally, how many people there are in total (population).

## What the fields mean

**Confidence level** tells you how often the method would produce a correct range if you repeated the survey many times. At 95%, in 95 out of 100 repeated surveys the true value would fall within the calculated margin. 95% is the usual choice; 90% suits quick, rough checks, and 99% decisions where a wrong call is costly.

**Margin of error** is the allowed deviation in percentage points. If a survey shows 62% of customers are satisfied, with a ±5 point margin the true value is, at the chosen confidence, between 57% and 67%.

**Population size** is the total number of people you draw conclusions about: all customers of a chain, all employees, all club members. The field is optional. For a large population, leave it empty — the result hardly changes. For a small one (a few hundred or a few thousand), enter it and the required sample will be smaller.

## How it is calculated

The calculator uses the standard formula for estimating a proportion:

- n₀ = z² · p(1 − p) / e², where z is the normal-distribution value for the chosen confidence (1.645 for 90%, 1.960 for 95%, 2.576 for 99%), e is the margin of error as a fraction (5% = 0.05), and p = 0.5.
- If a population N is entered, the finite population correction applies: n = n₀ / (1 + (n₀ − 1) / N).
- The result is rounded up, because half a respondent does not exist.

The value p = 0.5 is chosen on purpose: it gives the largest sample, so the calculated number is enough whatever the result turns out to be.

## Quick table for a large population

| Margin of error | 90% confidence | 95% confidence | 99% confidence |
| --- | --- | --- | --- |
| ±10 points | 68 | 97 | 166 |
| ±5 points | 271 | 385 | 664 |
| ±3 points | 752 | 1,068 | 1,844 |
| ±2 points | 1,691 | 2,401 | 4,147 |

The table shows an important pattern: when the margin of error is halved, the required sample grows about four times. Going from ±5 to ±2.5 points doesn't cost twice as many responses — it costs four times as many.

## How to choose the margin of error

The margin should be smaller than the difference that matters for your decision. If you will change opening hours only if more than half of customers want it, and you expect a result around 55%, a ±5 point margin won't give a clear answer. If you track a general satisfaction trend quarter by quarter, ±5 points is usually enough.

A practical rule: ±5 points for most business surveys, ±3 points when you compare results over time and expect small changes, ±10 points for a quick check before more detailed research.

## A sample is more than a number

The formula assumes every member of the population had an equal chance of being selected. In practice this is rarely fully true. If only customers who scanned a QR code at the checkout answer, the results describe in-store shoppers who like to give feedback — not those who order online.

So before calculating the number, decide who gets invited: every loyalty-card customer, every tenth receipt, all employees. A good selection of respondents matters more than a few hundred extra responses.

## From the number to the survey

Once you know how many responses you need, divide that by the expected response rate to find out how many people to invite. For a checkout survey, estimate how many customers pass by each day and how many days the survey should run. With the free Fidestat Forms plan you can build the survey, share it by QR code, link or email, and track how many responses have arrived against your target.

## Examples

- A pharmacy chain wants to know what share of customers are satisfied with pharmacist advice. There are many customers and the goal is ±5 points at 95% confidence: 385 responses are needed.
- A company with 2,000 employees runs an engagement survey. With a population of 2,000, 95% and ±5 points, 323 responses are enough instead of 385.
- A gym has 300 active members and wants ±5 points: it needs 169 responses — more than half of its members. With small populations the sample quickly becomes a large share of the whole.
- A marketing team tests two messages and wants ±5 points for each: that is 385 responses per message, 770 in total.

## Common mistakes

- Counting invitations sent instead of responses received. The calculator gives the number of responses; you need to invite more people.
- Reporting the whole-sample margin of error for subgroups. If only 60 of 385 respondents are from one city, the margin for that city is about ±12.7 points.
- Believing a large sample fixes a biased selection. A thousand responses from your most loyal customers still describe only your most loyal customers.
- Entering the margin of error as a fraction (0.05) instead of a percentage (5).
- Leaving out the population size when it is small (a team, a department, a club) and planning for more responses than you need.

## Frequently asked questions

### How many respondents are enough for a survey?

For a large population, 95% confidence and a ±5 percentage point margin of error, you need 385 responses. For ±3 points you need 1,068, and for ±10 points 97. If the population is small (say, 300 employees), you need fewer — enter it in the “Population size” field.

### Why does the calculator use p = 0.5?

We don't know the result in advance, and p = 0.5 gives the largest possible sample. That way the calculated number is enough no matter how people answer.

### Do I need a bigger sample for a huge population?

No. Above a few tens of thousands of people, population size barely affects the sample you need. For a town of 50,000 and a country of 6.6 million the number is practically the same: about 385 for ±5 points.

### How many invitations should I send to get the responses I need?

Divide the required number of responses by your expected response rate. If you need 385 responses and expect 20% of invitees to answer, invite about 1,925 people.

### Do the results hold if the sample isn't random?

The formula assumes a random sample. If only people who chose to answer take part (for example, a QR code at the checkout), the results describe those respondents well, but the margin of error does not cover selection bias. Who gets invited matters as much as how many answer.

### Do I need a bigger sample to compare groups?

Yes. The margin of error applies to the whole sample. If you need reliable results for each group separately (for example men and women, or three stores), calculate the required number for each group.

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