Automation A/B Test & Explorer: test and analyze your scenarios
You can now test several versions of the same automation to compare which one performs best, thanks to the Percentage split controller. This lets you steer your loyalty scenarios based on real data rather than intuition.
The A/B test: the "Percentage split" controller
The Percentage split controller distributes the customers entering your automation across several branches. Each branch is a variant (A / B / C) that you can build differently — for example two distinct re-engagement offers — so you can compare which one performs best.
Add the "Percentage split" controller at the point in your scenario where you want to split customers into several paths.
Create your variants — each branch (A, B, C…) receives a share of the customers and can contain its own steps and actions.
Set the percentage of each branch manually, based on the share of customers you want to send to each variant.
Use the "Equalize" button to automatically split the branches into equal parts in one click.
💡 Make sure the sum of your branch percentages matches the distribution you want and equals 100%. The "Equalize" button is the fastest way to reset to an even split.
Moving to "test & learn"
The split controller lets you compare several variants of a scenario on real data. You split your customers across two or three versions of an automation, then, to read each variant's results (how many customers move all the way through each branch), you use the automation's Explorer mode. Once you have identified the best variant, you can roll it out to all your customers.
💡 Concrete example:
You have a reactivation scenario running but you don't know whether it performs. You add a "Percentage split" controller right after the trigger, at 50/50: variant A sends an immediate follow-up, variant B waits 3 days before following up. A week later, you switch to Explorer mode: you see that branch B loses half of its customers on the delay step, while branch A goes all the way. The "should we wait or follow up right away?" debate is then settled with numbers — and you pinpoint the exact step where your customers drop off.
To analyze your variants' results step by step, see the Explorer section of our Discover automations guide.
Going further
To understand the full structure of an automation (triggers → controllers → actions), see the related articles:
If you have any doubt about setting up an A/B test, contact our support team via the chat.