Learn and master Clay with Clay university
Featured in Clay University
A/B Testing Your Message
Open A/B test in the campaign editor and Sequencer spins up a second variant. If you want to cancel the test, click through, select the sequence you want to keep, and it reverts.
Once you have your two variants, they can differ in the subject line, the body, or both. Click test body and you can edit that variant's copy directly. Add more messages to the sequence and you can add testing on those too, and if you no longer want a test on a given message, reset it.
☝️ Change One Thing
Say variant A has a different subject line, a different opening, a different call to action, and a whole different positioning, and it wins. All you've learned is that variant A is better. You have not learned anything you can actually use again, so on the next campaign you're back to guessing.
The more things you change between the two variants, the harder it is to understand what actually made the difference.
So pick one variable. Test the subject line this week, keep everything else identical, and when one variant does better you know the reason why. Next week, test the opening sentence.
📊 Reading the Compare
Switch to a campaign that has already run and open compare.
In the example in the video, variant A uses spintax in the subject line and an AI-written opening sentence, variant B uses neither, and everything else is word for word identical.
The view splits the analytics per variant, and you get:
- Emails processed and emails complete
- Conversion
- Sequence statistics
- A summary at the bottom that gives you the short version
On this campaign, variant A came back with a 1% reply rate and variant B with 0.6%. If reply rate is the number you're optimizing for, that points at continuing what variant A did.
Treat that as a direction, not a verdict. It's one test, on one campaign, asking one question.
There's plenty of other data worth monitoring depending on what you're trying to learn, bounces and unsubscribes among them. What you optimize for is up to you and how you're thinking about the experiment.
⚠️ What This Doesn't Do
Worth flagging plainly, because it changes how you should read the compare:
- There is no confidence threshold
- There is no minimum sample size
- Nothing here will tell you whether a result is significant, because Sequencer isn't calculating statistical significance yet
So the workflow is: launch both, leave it a week or two, look at the numbers, and make a judgment call. Nothing is hidden from you, but nothing is doing the statistics for you either. If you want real statistical rigor, export the numbers and do the math yourself.
🏆 Rolling Out the Winner
Once you've decided which variant won, roll that winner out and everyone still to be sequenced gets that email.
One habit worth building alongside it: write down what you tested and what happened, somewhere outside this tool. Six or seven campaigns from now you'll want to know whether you already tried the short subject line, and by then the campaign that ran it will have been archived for months.
🚀 What's Next
You can now set up an A/B test, read the compare view across both variants, roll out a winner, and size up how much weight a result actually deserves.
Next we'll build the full sequence around the message you've been writing, and run it through the spam checker.
