Skip to main content

Generate test ideas

PBX Ideate transforms identified friction points into specific, actionable optimization hypotheses and test ideas.

Generate hypotheses​

The system generates a hypothesis—a theory of change that articulates how fixing the root cause will improve the user outcome.

Experiments and variations​

For each page you analyze, the system generates one or more hypotheses. Each hypothesis includes specific test ideas that are ready to be implemented directly in the PBX interface.

To help you decide which tests to launch first, every idea is assigned an Impact Score. Powered by Confidence AI, the Impact Score represents the likelihood of the test producing a meaningful conversion lift, allowing you to focus on high-impact opportunities.

Each experiment suggests concrete variations—specific changes to text, layout, or styling—so you can start testing immediately without needing to design from scratch.

Scanned pages​

To generate test ideas, add specific URLs to the Scanned pages tab.

  1. Navigate to Insights > Test Ideas.
  2. Click Add pages. The Add pages to scan pop-up opens.
  3. Select the project for the page.
  4. Enter the page URL.
  5. Click the + icon > Add pages.

You can analyze a maximum of 10 scanned pages per project. To check the status of pages across projects, click View details.

note

Hypothesis generation is not instant. Wait for the analysis to complete, which takes approximately 10–15 minutes after you scan a page.

View and manage test ideas​

You can access AI-generated test ideas directly within Kameleoon.

Select pages for analysis​

  1. Sign in to Kameleoon and navigate to Insights > Test Ideas.
  2. From the dropdown menu, select the website to analyze.
  3. Add the specific URLs to analyze.
note

The AI analysis process is thorough and may take up to 15 minutes to complete.

Analysis quotas and strategy​

You can analyze a maximum of 10 pages per project within a 30-day period. Once a page is analyzed, a new analysis for that URL is unavailable until the 30-day period expires.

Because of this limit, select page categories rather than multiple similar pages.

  • Recommendation: Analyze one representative URL for a category (for example, one product detail page) rather than adding 10 different product pages.
  • Logic: Usability friction identified on one product page is likely present on all similar pages.
  • Targeting: Although the analysis is performed on a single URL, apply the resulting experiment to all pages in that category (for example, targeting all product pages) using standard targeting conditions in PBX.

Review suggestions​

When a page is expanded, the system displays the generated hypotheses. For each hypothesis, the system lists specific, actionable test ideas.

Each idea card displays:

  • Test name: A concise title for the experiment.
  • Impact Score: A predictive score (0-10) estimating the likelihood of this test producing a conversion lift. This score is derived from analyzing thousands of historical test results to identify factors that predict success:
    • 8-10 (High confidence): Strong evidence, proven patterns, and clear mechanisms.
    • 5-7 (Moderate confidence): Reasonable hypothesis with partially supported patterns.
    • 1-4 (Low confidence): Speculative mechanism or unsupported patterns.

Analyze details​

To understand the logic behind a suggestion:

  1. Click the three dots menu next to a specific test idea.
  2. Select See details.

OR

  • Click See diagnosis.

A pop-up panel displays details about the test. The app highlights the targeted element at the top of the panel.

Hypothesis​

Under the Hypothesis dropdown, you can see:

  • Levers: Shows the Conversion Master Lever, Lever, and Sub-Lever that form the hypothesis' basis.
  • Insight: Provides a logical synthesis of why the change should work.
  • UX Observations: Lists specific elements on your page that triggered the recommendation.
  • Best Practices: Cites behavioral principles derived from academic research and successful test outcomes, ensuring recommendations are grounded in real-world performance data.

Best practices include:

  • Attention Follows Visual Weight: Universal perceptual law validated across hundreds of tests.
  • Explicitly Framing Reduces Interpretation Effort: Mechanism-aligned pattern from clarity testing.
  • Primary CTAs Require Dominant Contrast: Design heuristic with consistent positive results.

Test Idea​

The Test Idea dropdown shows the specific variations proposed for that test idea. These variations represent concrete ways to implement the test idea (for example, changing the checkout flow to a single-page layout).

Launch an experiment​

Once a promising idea is identified, move directly to the implementation phase.

  1. In the sidebar, select the variation to test.
  2. Click Create in PBX.
note

If you've already created a variation, the CTA changes to View in PBX.

How experiments are created​

  • First variation: The first time you click Create in PBX for a test idea, Kameleoon creates a new experiment containing the selected variation.
  • Additional variations: To test multiple variations against each other (for example, Variation A versus Variation B), select a second variation from the list and click Create in PBX again. The system automatically adds the new variation to the experiment created in the previous step.

The PBX interface opens with the targeted element and suggested prompt ready for review. You can then finalize the design and launch the experiment.