How to Start a Fresh A/B Test

Answer

When you want to run a new A/B test with clean metrics, you have two main options: reset assignments on an existing experiment or create a new campaign. Which approach to use depends on what you are testing and whether you want to preserve historical data.

Option 1: Reset assignments

Best when you want to reuse the same campaign and paywall setup but start measuring from scratch.

1. Navigate to your campaign, select the audience, and click the Paywalls tab to view the experiment group.
2. When editing presentation percentages, look for the refresh icon (circular arrow) next to where it says "X assigned" below any paywall variant.
3. Choose to reset a single variant to clear history for just that paywall, or reset all paywalls to clear history for every variant in the experiment.

Resetting creates a new variant ID behind the scenes. All analytics for that variant, including conversions, ARPU, users, and trial metrics, start fresh from that point forward. The paywall configuration itself stays the same.

Option 2: Create a new campaign

Best when you want to keep your previous test results intact for reference, or when you are testing a fundamentally different audience or placement.

Create a new campaign with the paywalls you want to compare. This gives you a completely separate experiment with its own analytics, while your old campaign's data remains untouched.

How assignments work

When a user matches an audience and is assigned a paywall variant, that assignment is sticky. The user will continue to see the same variant regardless of percentage changes. Changing presentation percentages only affects new users.

Resetting assignments clears these sticky assignments, so all users (new and returning) will be reassigned based on the current percentages. A new campaign has no prior assignments, so all users start fresh.

Which option to choose

• Reset assignments when you want to keep your campaign structure and do not need to reference old test data.
• Create a new campaign when you want to preserve previous results or are changing the audience, placement, or overall test setup.