This feature is in BETA version. For more information, contact your Customer Success Representative.
Real-Time A/B Testing allows marketers to experiment and optimize customer journeys by testing multiple variants of a campaign treatment—such as message content, timing, or delivery channel—in real time.
With A/B Testing in Treasure Data Real-Time Journeys, you can test what works best and continuously improve performance without waiting for batch updates.
Add A/B Test steps anywhere in the journey (except after end or jump nodes).
Create up to 8 variants plus an optional control group.
Customize test and variant names (default: A–H).
Split traffic equally or assign custom percentages.
Supports nested tests (up to 10 layers) and merge branches post-test.
Users are randomly and consistently assigned to a variant when they reach the A/B test node.
Variant assignment is based on a stable internal ID, ensuring consistent treatment across sessions and re-entries.
Variant assignments are stored and can be referenced for downstream analytics and targeting.
Even after users exit a journey, their test membership is retained for future re-targeting or analysis.
The following enhancements are planned for an upcoming milestone:
Marketers will be able to pass A/B test and variant data to external systems during activation. This will support downstream personalization and performance measurement.
Activation payloads will include the test name and variant (if assigned).
If multiple A/B tests are nested, the system will send the closest relevant variant.
In the Real-Time Segment Editor, marketers will be able to include or exclude profiles based on their A/B test group membership.
- Support for targeting current or historical variant members.
- Useful for follow-up actions, retargeting, or suppressing specific variants.
Edit Protection: Once a journey is launched, A/B tests within it cannot be changed.
Delete Protection: A journey cannot be deleted if one of its A/B tests is referenced in another journey. This prevents accidental breakage of dependent experiments.
Campaign Objective: Optimize email subject lines
Stage 1: Add an A/B Test node testing two subject lines (Variant A and Variant B).
Stage 2 (Future): Use a drag-on rule to only show a follow-up offer to users from Variant A who opened the email.
Analysis (Future): Export variant membership through activation metadata to evaluate click-through and conversion rates.