Subject line A/B test

4 min read61
Copy for AI:

An A/B test sends two different subject lines to a small slice of the audience and, after a wait, sends the one that worked better to everyone else. You do not have to guess which phrase makes your base click: the campaign finds out and uses the answer.

In the example, Renata, from Casa Moura CafΓ©s, cannot decide for the 2026 harvest launch between "The 2026 harvest is here" and "{{first_name|cliente}}, the 2026 harvest has arrived".

Turn the test on

In the campaign editor, Content tab, right below the subject, turn on Test two subject lines (A/B). The subject already in the field above is A. Fill in:

  • Subject B β€” required and different from A (if they are the same, there is nothing to test). Accepts the same variables.
  • Preheader B β€” optional. Left blank, variant B goes out with the same preheader as A, and the test compares only the subject. Fill it in when you also want to test a different snippet.
  • Test slice β€” how much of the audience gets the test, split half and half between A and B. With 20%, 10% get A, 10% get B, and 80% wait for the winner.
  • Wait before deciding β€” how long after the slice goes out the winner is chosen. By click, 3 hours already pick the final winner in most cases.

If you switch the toggle off to compare and back on, subject B's text is still there. What counts is the toggle when you save.

The test covers only the subject and the preheader: the email body is the same for both variants. It is also not available in a campaign for spaces.

Before sending: the sample size

In the Send campaign window, the test summary says how many people get each subject now and in how many hours the rest goes out. The math uses who will receive it, already without blocked and unsubscribed contacts.

If each subject ends up with fewer than 500 people, the notice turns yellow: below that, the difference between A and B is noise, and the result is an impression, not a measurement. With a small base, raise the test slice or turn the test off and send a single subject.

How the winner is chosen

The subject with the higher click rate wins: how many people clicked out of how many received that variant. A tie keeps A β€” including zero to zero.

Opens do not decide, on purpose. Some email apps load images on their own, without anyone reading anything, which inflates opens for any subject. A click only happens through a person's action.

At the set time, the rest of the audience gets the winner automatically. You do not need to be on the screen.

Follow the test

On the campaign screen, above the tabs, the Subject line A/B test box shows the test rule and both variants side by side. While the campaign is Sending, the badge reads Testing Β· decides at and the time. Each variant shows the click rate, highlighted, plus Sent, Delivered, Opened and Clicked.

After the decision, the badge turns into Winner: A (or B), with by clicks or picked by hand, and the winning variant is marked.

In the Recipients tab, whoever received it in the test slice has the variant letter next to the address. Those who got the winner later have no letter.

Decide early

While the test is running, Decide now offers three options: By clicks so far (a tie keeps A), or send A, or send B, picked by hand. When you confirm with Decide and send the rest, the rest of the audience goes out right after with the chosen subject. This cannot be undone.

Use it when the difference is already clear and you do not want to wait, or when you know one of the subjects has a problem. Deciding is only possible while the campaign is sending, and only Manager and Administrator see the button.

Repeat the test

When you clone a campaign, the test settings come along (subject B, slice and wait), but the winner and the numbers stay with the original. So Renata can reuse the format in the next launch and test another phrase.

Where to go next

  • Create a campaign β€” subject, preheader and variables.
  • Review and send a campaign β€” the confirmation and scheduling.
  • Track a campaign's results β€” the full report.
  • Reports β€” the account's best time to send.

In short: turn the test on, write a different subject B, choose slice and wait, check in the confirmation that each subject will reach enough people, and let the click rate pick the winner β€” or decide early, knowing there is no going back.

Was this article helpful?