Run an A/B test
Send two or more versions of a campaign to part of your list, pick a winner by opens, clicks or by hand, and send it to everyone else.
An A/B test sends different versions of an email to a small part of your list. After a wait, the best version goes to everyone else. Each version can have its own subject line, sender and design.
Create an A/B test
- In the left sidebar, click Campaigns, then Create campaign.
- On the A/B Test card, click Create.
The A/B wizard has three steps, Campaign, Variants and Winner Rules, followed by Review & Launch.
Step 1: Campaign
- Enter a Campaign name.
- Choose the Audience list.
- Set the Test audience size: how much of the list gets the test versions. The rest gets the winner.
- Choose an Assignment strategy:
- Exact split: accurate to one subscriber. Best for most lists.
- Hash split: approximate, but faster on very large lists.
Step 2: Variants
- For each variant, set a Variant name, an Email subject, the From name, From email and Reply-to, and choose a template.
- Click Add variant to add another version.
- Set the Split for each variant. The splits must add up to exactly 100%.
To test only the subject line, give every variant the same design and change just the subject.
Step 3: Winner Rules
- Choose how the winner is picked:
- Highest open rate
- Highest click rate
- Manual selection: you look at the results and choose.
- Set Wait before deciding: how many hours after the test sends before the winner is chosen.
Launch and pick the winner
- On Review & Launch, check everything and click Launch A/B Test.
- The test versions go out first. The campaign shows Awaiting winner while it collects results.
- After the wait, SendCoop sends the best version to the rest of the audience. With Manual selection, open the A/B Test Report and click Choose as winner on the version you want.
Tips
- Test one thing at a time, such as the subject line, so you know what made the difference.
- Clicks are a more reliable measure than opens, because some email apps load images automatically.