Skip to main content
Cloning an experiment gives you a fresh draft that starts identical to the original, same variants, same targeting, same goals, with three important differences.

​
How to clone

From any experiment’s detail page or from the row menu on the experiments list, click Clone (or Duplicate). The dashboard opens the new draft immediately so you can rename it and adjust before starting.

​
What is different in the clone

New bucketing salt. The clone gets its own random salt so visitors are re-bucketed independently. A visitor who saw Variant B in the original does not automatically see Variant B in the clone, they are re-assigned from scratch. New experiment key code. The ABTESTLY_EXP_KEY is regenerated so abtestly.activate('key'), onApply('key', …), and any code that targets the original by key does not accidentally target the clone too. Fresh status. The clone starts as Draft, regardless of the original’s state.

​
What carries over

  • Every variant’s JS, CSS, and weights.
  • Traffic allocation and split.
  • Locations and Audiences (references, not copies, see below).
  • Activation mode and target selector.
  • Goals attached (as shared references, see below).
  • Schedule fields (start, end, max days) as cleared, so the clone is not immediately scheduled.

​
Goals become shared

If the original had inline goals, goals defined only for that experiment, cloning promotes them into the site’s goal library and attaches both experiments to the shared reference. The dashboard shows a toast telling you this happened. Editing the shared goal changes it for both experiments. If you want one to diverge, Clone to edit the goal from the library, then attach the copy to the experiment you want to differ.

​
Mutual exclusion groups do not copy

If the original was a member of an exclusion group, the clone is not added to the group. Exclusion group membership is a deliberate, intentional decision, cloning does not silently expand a group. Add the clone to the group manually if that is what you want.

​
When to clone

  • You want to run the “same test” on a different Audience. Clone, swap the Audience, ship.
  • You want to re-run a paused winner on more traffic. Clone the winner, bump allocation, ship.
  • You want to iterate a losing test with tweaks. Clone, edit the variant, ship. The old data stays with the original.

​
When not to clone

Cloning does not carry the original’s data. If you want the results page to keep counting the original’s exposures under the tweaked variant, do not clone, either Reset data on the original after editing, or leave the original alone and read its results as-is.