Skip to main content
An A/B test sends several versions of a campaign to a sample of the list, waits, picks the winner on the criterion you chose, then sends it to the rest of the list, on its own.

Switch the test on

The campaign’s Settings tab, the A/B test section:
2 to 5
default:"2"
Each version gets the same share of the sample.
5 % to 80 %
The share of the list that receives the test, drawn at random. With 20 % of a list of 10,000 people and two versions: 1,000 receive A, 1,000 receive B, and 8,000 will receive the winner.
criterion
  • Most unique opens: for testing subjects. - Most unique clicks: for testing content. - Most money earned: for testing what sells (you need to report your sales).
5 minutes to 7 days
Give your readers time: 2 to 4 hours for opens, a day for money.

Writing the versions

In the Content tab, Version A, Version B… buttons replace the language tabs. Each version has its own subject and content; the editor, the template and the settings are shared. A field left empty falls back to the campaign’s own.
Change one thing at a time. Two different subjects with the same content teach you which subject works better; two entirely different e-mails teach you nothing you can use again.

While the test runs

  1. Send now sends the sample, split between the versions.
  2. The scheduler checks every minute whether the wait is over.
  3. The version with the best score becomes the winner; on a tie, the first one (A) wins.
  4. The winner goes to the remaining recipients, at the same sending pace.
The Delivery tab repeats the test (“Both versions go to X % of the recipients…”) and, once it is decided, which version won and when.

Reading the results

In Statistics, the A/B test section shows each version with its subject, how many were sent, its opens and its clicks; the winner is marked Winner. In the Outbox, the Version column says what each person received.
A version already sent cannot be deleted, and the versions of an A/B test are not translated. To test in several languages, make one campaign per language segment.