A/B test a subject line
Try two subjects on a slice, then send the winner to the rest.
Turn on A/B in the campaign editor and give a second subject line. A slice of your audience gets each, and after a wait you set, the winner goes to everyone else.
How it works
- Choose what share of the audience is the test — the rest are held back.
- The slice is split randomly and evenly, and each half gets one subject.
- After your wait, the winner is picked and sent to everyone remaining.
The held-back people genuinely aren’t queued yet, so nothing can leak to them early.
Sizing it
Below about 50 per variant, you're reading noise
With 25 people each, a difference of two clicks looks decisive and means nothing. The editor warns when your split lands there.
The trade-off: a bigger test is more reliable but leaves fewer people to benefit from the winner. On a list of a few hundred, consider testing across two campaigns instead — same audience, different subject, compare afterwards.
Decide on clicks
Clicks is the default, and it’s the right one even though you’re testing subject lines. Apple opens messages on recipients’ behalf, so an opens-based winner is close to a coin flip.
A subject line’s job is getting someone to engage, and a click is the honest evidence they did.
How this differs from an in-flow test
This one decides on a clock: a campaign’s test slice is spent the moment it sends, so there’s a remainder waiting and every hour you wait is an hour it isn’t landing.
An in-flow test decides on a sample size, because a flow never stops enrolling people and waiting costs nothing. Same idea, opposite stopping rule — which is why they’re separate features.
Common issues
The campaign is stuck “sending”
Expected during the wait — the test has gone, the remainder hasn’t. It completes when the winner rolls out.
The remainder never went
Something failed the re-check at rollout time — usually credit. Top up and it goes on the next pass; nothing is stranded.
Both variants performed identically
A real result: those subject lines don’t differ in a way your audience cares about. Test something bolder — a different angle, not a reworded version of the same one.
FAQs
Can I test the email content, not just the subject?+
Not on a campaign — it tests subject and preheader, with one design. To test two different emails, use an in-flow split.
What happens on a tie?+
Variant A. Something has to win, and a coin flip would be worse.
How long should the wait be?+
A few hours suits most lists — long enough for a real spread of engagement, short enough that the remainder still arrives on the day you intended.