Declare and adopt an A/B test winner

Apply a winning variant’s configuration to your control directly from the test page.

Adopting an A/B test winner applies your winning variant’s configuration directly to your production search, eliminating the need to manually copy settings or reconfigure your index.

This action is available only when the declared winner is a variant (not the control).

Before you begin

You’ll need:

• Admin access to your Algolia dashboard to change index settings and promote configurations

  Users without admin access can declare a winner but can’t promote settings to modify index configurations. Full admin access is required for the complete adoption process.

• An A/B test with at least 80% completion so that the option to declare a winner is available

• Understanding of your current index configuration to verify changes are working properly

• Ready to promote the settings of your winning test variant to the control index. The entire process takes 5 to 10 minutes for the technical adoption and 2 to 3 hours of monitoring

Declare A/B test winner

1. Go to the Algolia dashboard and select your Algolia application.
2. On the left sidebar, select Search.
3. On the A/B testing page, select your A/B test you want to process.
4. On the A/B test detail page, select the variant you want to choose as winner and click Declare winner.

Your A/B test must be at least 80% complete (based on its scheduled duration) before you can declare a winner. After declaring a winner, you can immediately adopt the winning configuration. (Declaring a winner itself doesn’t change your indices).

Promote settings

After declaring a winner, click Promote settings to copy the winning configuration to your control.

Adoption window: You have 14 days after your A/B test ends or you stopped it to adopt the winning configuration. After this period, you’ll need to apply the settings manually to your index.

Confirm promotion

To confirm updating the index settings:

• If the winning variant uses the same index as the control: Enter COPY
• If the winning variant uses a different index as the control: Enter UPDATE

Click Confirm to proceed.

Configuration changes during adoption

The adoption action copies most configuration settings from your winning variant to your control, but there are some important limitations to be aware of.

What gets applied

• Update the control’s index settings to match the winning variant (or custom search parameters if variant index is the same as the control’s).
• Toggle feature settings on or off to match the winner (limited to):
  • Dynamic Re-ranking
  • NeuralSearch
  • Personalization
  • Advanced Personalization

What doesn’t get modified

• Records: Your actual data is never modified.
• Synonyms and rules: These remain unchanged during adoption.
• Replica limitations: If the control is a replica (virtual or standard), the system applies only the subset of index settings listed here during adoption.

Verify adoption

After promoting settings, verify the changes work as expected:

1. Check your index configuration in the dashboard to confirm your control index received the winning variant’s settings.
2. Check your analytics for any unexpected changes in the next 2 to 3 hours.

Revert settings

You have a 15-day window after your A/B test ends or you stopped it to revert the adoption.

To revert to your previous settings:

1. Open the A/B test details page in the dashboard.
2. On the control variant, click Revert.

1. Confirm by entering REVERT and clicking Confirm.

2. The original winning variant still appears as the winner, but no longer shows as “promoted”.

Important reversion details:

• After reverting, you won’t be able to promote settings again for this A/B test.
• The same limitations that applied during initial adoption also apply when reverting (replica restrictions, feature settings constraints).
• Only the settings that were originally applied during adoption can be reverted.

Need help? If you experience unexpected behavior after reverting settings, contact Algolia support.

See also

• Create and run an A/B test
• How to read your A/B test results