---
type: "Glossary Term"
title: "What is an A/B test? | Roicool Glossary"
description: "What is an A/B test? Comparing two versions on randomly split audiences to see which performs better: how it works and what to watch."
url: "https://www.roicool.com/en/glossary/a-b-test"
lang: "en"
dateModified: "2026-10-20"
resource: "https://www.roicool.com/en/glossary/a-b-test"
alternates:
  tr: "https://www.roicool.com/sozluk/a-b-testi.md"
generated: { by: "process:roicool-cms", at: "2026-10-20T15:58:40.406Z" }
related:
  - { title: "Call-to-Action", resource: "https://www.roicool.com/en/glossary/cta.md" }
  - { title: "a Landing Page", resource: "https://www.roicool.com/en/glossary/landing-page.md" }
  - { title: "User Experience", resource: "https://www.roicool.com/en/glossary/user-experience-ux.md" }
  - { title: "Conversion optimization", resource: "https://www.roicool.com/en/services/conversion-rate-optimization.md" }
  - { title: "Glossary", resource: "https://www.roicool.com/en/glossary.md" }
---

# What is A/B test?

An A/B test shows two versions of a page, ad or email to randomly split audiences in the same period and measures which one performs better on a chosen goal, such as clicks, form fills or sales. The result is read only once enough data shows the difference is not down to chance.

## In short

An A/B test is the standard way to settle design and copy decisions with observed behaviour instead of opinion. Marketers use it on single variables such as a landing page headline, an ad image, form length or an email subject line. The goal metric and required sample size are fixed before launch, and the test runs for its planned duration. For example, a B2B site might compare a "Book a demo" button against "Get a quote" on form submission rate over two weeks.

## Key points

- Visitors are randomly split: group A sees the control, group B the changed version.
- Changing one element per test makes the cause of any difference clear.
- Decide the goal metric and test duration before the test starts.
- With low traffic, small differences may be noise, so avoid stopping early.
- Campaigns, seasonality and weekdays can skew results.

## How does an A/B test work?

Visitors are split at random: group A sees the current version (control), group B the changed one. The same goal metric is compared over the same period. Changing one thing per test (headline, offer, form, image) makes the cause of a difference clear.

## What to watch

- **Sample size:** with low traffic, small differences may be noise; work out the duration before you start.
- **Don't stop early:** results swing in the first days.
- **One goal metric:** decide what you measure before the test.
- **Seasonality:** campaigns and weekdays can skew results.

## Why it matters

A/B testing grounds design and copy decisions in data rather than opinion.

## Related terms

- [Call-to-Action](https://www.roicool.com/en/glossary/cta)
- [a Landing Page](https://www.roicool.com/en/glossary/landing-page)
- [User Experience](https://www.roicool.com/en/glossary/user-experience-ux)

## Related service

- [Conversion optimization](https://www.roicool.com/en/services/conversion-rate-optimization)

Breadcrumb: [Home](https://www.roicool.com/en) › [Glossary](https://www.roicool.com/en/glossary) › A/B test

Related pages:

- [Glossary](https://www.roicool.com/en/glossary)

Source: https://www.roicool.com/en/glossary/a-b-test · Last updated: 2026-10-20