A/B testing

Stop guessing which email performs best. Create competing variants, send them to a slice of your audience, and let EmailFlow AI measure the results and automatically deliver the winner to everyone else.

What you can test

An A/B test compares two or more variants of the same campaign. Each variant is a full email, so they can differ in whatever you want to learn about:

Subject line — the most common test, and the biggest lever on open rate.
Sender name — does a person or the brand get more opens?
Content — different layouts, copy, images, or calls to action.

You assign each variant a share of the test audience, and you can mark one as the control.

How winner selection works

Choose how the winner is decided and EmailFlow AI does the rest:

By open rate — best for subject-line and sender tests.
By click rate — best for content and call-to-action tests.
Manual — you review the results and pick the winner yourself.

For automatic criteria, you set a wait window (anywhere from 1 to 720 hours). When the window closes, EmailFlow AI evaluates the variants and declares the winner.

Testing on a subset

You decide what portion of your audience receives the test — from a small sample up to the entire list. With a partial test (say, 20%), the variants go to the test slice first; the remaining contacts then receive the winning version automatically. Test the whole list when you simply want to split your audience evenly between variants.

science

For statistically meaningful results, test on a large enough sample and give the wait window enough time for opens and clicks to accumulate — a few hours at minimum, longer for smaller lists.

Auto-sending the winner

Once the winner is declared — automatically by your chosen criteria or manually by you — EmailFlow AI sends it to every remaining recipient who hasn't been emailed yet. You get the benefit of a controlled experiment without lifting a finger after setup. Results for every variant appear in your campaign reports.

Frequently asked questions

How many variants can I test? Two or more. Most tests compare two versions for clear results, but you can add more when you have several distinct ideas to compare.

What if it's a tie or there's not enough data? Give the test a long enough wait window and a large enough sample so opens and clicks can accumulate. For close calls, choose the manual criterion and review the numbers yourself before declaring a winner.

Does the winner reach everyone? Yes — once the winner is decided, it's sent automatically to every remaining recipient who wasn't part of the test slice, so no one is left out.