Skip to main content

Run an A/B Split Test End to End

A worked example of setting up, sending, and reading the results of an A/B split test.

What you'll build​

By the end of this tutorial you'll have run a single Broadcast that tests two subject-line variants against each other. A small slice of your subscribers receives each variant, the platform measures which performs better, and the winning variant is sent to everyone who remains. You'll finish by reading the results and knowing which version won and why.

Before you begin​

  • An Edition that is ready to send, saved in a Project Folder.
  • A subscriber audience large enough to split meaningfully — a few thousand or more gives clearer results than a few hundred.
  • A single, clear thing to test. Start with the subject line; test only one element per Broadcast so the result is unambiguous.
  • Permission to send in your account, and a sender/reply-to address already configured.

Steps​

  1. Open your Edition and start a new Broadcast from it, then choose the A/B or split-test option (confirm the exact steps in your account).
  2. Define Variant A and Variant B. For a subject-line test, write two distinct subject lines and keep everything else identical so the subject line is the only difference.
  3. Choose the test size — the percentage of subscribers who receive the test (for example, 10% to each variant, leaving 80% for the winner).
  4. Choose the winning metric — usually open rate for subject-line tests, or click rate when you're testing content or calls to action.
  5. Set the test duration — how long the platform waits before picking a winner. Give it long enough to gather signal (often a few hours).
  6. Select your audience, review the schedule, and launch the Broadcast.
  7. When the test window ends, the winning variant is sent automatically to the remaining subscribers.

Next steps​

  • Open the Broadcast's report and compare opens, clicks, and the declared winner across variants.
  • Note what won and roll the learning into your next Edition.
  • Try testing a different element next time — send time, preheader, or a call to action.

Canonical terms: Author, Edition, Folder (Project Folder), Broadcast. See the Glossary.