Action - A/B Test

The A/B Test action randomly splits people into two or more paths in a Flow. 

This action is best used if you want to A/B test a multi-step flow – for instance, sequences of emails. If you'd wish to run a standard single email A/B test (e.g., to figure out which subject line has a higher open rate) it's best to use the A/B test feature in Broadcasts (coming soon).

How to set up A/B Test action step?

Click on the step to configure it. By default, the A/B Test is split into two equal groups – Variant A and B. This means that 50% of the people (randomly chosen) will flow into Variant A and 50% will flow into Variant B. However, you can change this if you'd like a larger or a smaller portion to continue down a specific path.

Finally, you need to connect both the A and B links to another step:

Activate the Flow and let it run for a certain amount of time before you decide which path/sequence you'd like to keep. 

To see the performance of the emails in the flow, click on a specific email step and select the Report tab on the right. Note that the Report tab currently shows metrics from the last 30 days only (we're working on implementing a date picker):

Once you choose the winner variant, simply delete the A/B Test step and the loser path/sequence:

Creating an A/B Test with more than two variants

To add more variants to your A/B Test simply click on the "+ Add variant" button. Then, click on "Distribute % evenly" if you wish to create equal groups.

You can distribute the variants however you like. Note that the total sum of all of the variants must be 100%, otherwise, the A/B test will not work properly. 

If you wish to delete a variant click on the trash icon.

Finally, make sure to connect all of the variants to another step.

Example use cases

Randomly show different Intercom in-app messages

Randomly assign leads to two or more different salespeople

Use the Create HubSpot Deal action step and assign a different owner to each step:

Did this answer your question? Thanks for the feedback There was a problem submitting your feedback. Please try again later.