Work out how big a sample an A/B test needs

I want to launch a test tomorrow, and I have no idea how many visitors it actually needs to give a trustworthy answer.

Beginner to intermediateRead 5 minDoing: under an hour5 steps

What you'll have at the end

The number of visitors each variant needs, calculated before the test goes live, plus how many weeks of real traffic it will take to reach that number.

Before you start

You need

A specific page or feature you plan to A/B test, with at least a few weeks of real traffic history for the metric you're testing.

Not covered

Reading the test's results once it has actually run and deciding whether the difference you see is real.

Leans on

5 steps

Where to go next

Compare two groups' averages with a t-test

once the test has actually run and you have real numbers from both groups to compare.

An A/B test isn't significant, and the team still needs a call

if the test finishes without reaching significance and you still need to decide what to do next.

Checked 26 Aug 2026

Part of the Statistics Fundamentals cookbook