Create a test optimization

Make a traditional test or AI-optimized test to experiment with different page elements.

Test optimizations are comparable to standard A/B tests. You can create variations to test out different versions of the page to see which version(s) resonates most with your visitors. You can add several variations in one test (e.g., A/B/C/D).

Traditional test vs. AI-optimized test

Traditional test optimizations are like typical A/B tests. Your variations go live with a fixed percentage of randomly allocated traffic (e.g., A = 50% of site traffic, B = 50% of site traffic), where they'll go head-to-head over an extended period of time to find a single "winner." The winner is determined once the test reaches statistical significance and displays the top performer.

AI-optimized optimizations leverage machine learning to determine which variations perform best for different types of visitors. Traffic is dynamically allocated based on visitor engagement. A simplified example might be that variation A performs well for visitors who find your page from a search engine, whereas variation B performs best for visitors from a paid ad. Both variations work great for each visitor group, so both are "winners" for their respective groups, and traffic is allocated appropriately. If visitor behavior changes later on, the AI adapts traffic allocation based on what they're interested in.

About editing traditional tests

Traditional tests use statistical significance (stat. sig.) to measure results. Stat. sig. requires all variables to remain unchanged to ensure reliable data — which means the optimization must run from start to finish without edits. This keeps results stable and directly comparable to the control. How long that takes depends on your traffic and visitor behavior.

To account for this required stability while still allowing you to make updates, we use intervals. You can stop the test to make changes (e.g., add a variation, update the goal, or apply an audience). When you’re done, relaunch the test — this starts a new interval and resets the time needed to reach stat. sig.

For small fixes (e.g., correcting a typo), you technically can edit a live variation — but doing so introduces new variables that may affect your data. To preserve result integrity, we recommend stopping the optimization before making your updates.

How to create a test

To draft a test optimization:

1. Open the Pages panel
2. Open the page you want to optimize
3. Click the Optimizations dropdown in the top bar
4. Click New test
5. Choose either Traditional test or AI-optimized test
6. Click Create

Your optimization is now in a draft state and ready for you to configure it and create variations.

How to configure your test

These are optional settings you can apply to your test optimization.

Limit who can see the test

By default, every visitor who lands on a page where the optimization runs is eligible to see its variations. You can limit this pool by applying an audience — only visitors who match that audience will be eligible for the optimization.

Learn how to limit the optimization to an audience.

Adjust traffic allocation

In a traditional test, traffic is split evenly across all variations by default. You can change the allocation while the test is in a  draft or off state.

To change traffic allocation in traditional tests:

1. Open the Pages panel
2. Open the page where your optimization is saved
3. Click the Optimizations dropdown in the top bar
4. Click Settings & goals
5. Click Traffic allocation
6. Choose Manual
7. Adjust the "Split" percentage per variation
8. Click Save when the total split percentages equal 100%

Next steps for optimizing

Now that you have the optimization configured, it's time to create variations.

• Build out your changes in variations
• Launch your variations to your visitors
• Review the results after some time has passed