A/B testing
Last edited: 2025/03/17
A/B testing helps you identify the optimal parameters for your setup by analyzing how different settings impact performance.
In the Yieldbird Platform, you can run A/B tests on various Domain-level parameters, such as Prebid Bidders (On/Off), User ID Modules, Lazy Loading settings, Refresher configurations, and more.
Setting up A/B test
- Go to A/B testing tab, then click ‘Add new’
- Name the test and select the Domain on which you want to conduct the test
- Select Parameters to Test Choose the parameters you want to test. Multiple selections are allowed.
- Define Variant settings (Domain configuration Variants)
- You can add up to 10 Variants in a single test. Click ‘Add New’ to include additional variants.
- Variants cannot be renamed, but you can modify the Reported key value (the identifier for traffic assigned to a Variant). The name must not contain spaces or capital letters.
- In the Variants table, you can only modify settings for the parameters selected in the Parameters section. To add more parameters, return to the Parameters section above.
- Choose an Action:
- ‘Save and Start Test’ – The test will begin immediately.
- ‘Save as Draft’ – The test configuration will be saved, allowing you to start it later.
You can have only one A/B test running per Domain at a time.
If you create a new A/B test on a Domain where one is already active, you won’t be able to start it immediately. Instead, you can save it as a draft, which can only be started once the ongoing test is completed.
Monitoring A/B test results
You have two options for tracking A/B test results:
1. GAM Reporting
Track A/B test results directly in Google Ad Manager (GAM) by generating reports based on the key-value dimension. The key-value to look for is "qw_ver".
2. Prebid Analytics
View A/B test results in the Yieldbird Platform under the Prebid Analytics tab (Prebid data only).
To access this:
- Go to: Performance > Prebid Stack > Analytics > A/B Tests.
Finishing A/B test
Good practice: We recommend to make each A/B test running for at least full 7 days to catch all the weekly seasonality.