Skip to main content

A/B test types

The current builder has one experiment model: compare 2–9 complete videos against one supported goal. It does not offer named hook, thumbnail, CTA, player-skin, localization, or multivariate test types.

Supported goals​

Choose one goal for the entire test:

  • CTA clicks
  • Conversions
  • Retention
  • Watch time
  • Lead capture

The results page formats the corresponding observed rate for each video variant.

Variants and control​

Every variant is a complete video selected from the workspace. Choose exactly one control. Use Equal split or Manual split; manual percentages must total 100%. You can add, remove, or replace videos while the test is a draft. After launch, videos are locked and only traffic allocation remains editable.

What you can test​

To test a hook, CTA, thumbnail, pitch, player treatment, or localization idea, prepare each complete video version first and then compare those videos. The builder does not edit individual components or construct a matrix of combinations.

Promotion​

Promote any selected variant manually, or enable the per-test time-based auto-promotion option. Auto-promotion uses the current goal-rate leader after the configured delay; it is not a confidence-based or guardrail-based workflow.

Best practices​

  • Change one major idea between complete-video variants when you want an interpretable comparison.
  • Match the goal to the change you made.
  • Keep exactly one control and verify the split totals 100%.
  • Use a new or duplicated draft when you need to change videos after launch.

Frequently asked questions​

Which test types can I select?

There is no named-type selector. Current tests compare 2–9 complete videos against one goal.

Can I test only a CTA or thumbnail?

Create complete video versions containing those alternatives, then select the videos as variants. The test builder does not edit components.

Can I build a multivariate matrix?

No. The current builder has no cell-matrix workflow; see Guardrails, SRM & multivariate.