Problem
In its current state, xLab doesn’t allow experiment owners to sample test subjects based on their login status. Users are thus not able to configure experiments that collect metrics from anonymous users.
Suggested solution
To enable experiment owners to benefit from the new anonymous testing capabilities, we’ll provide them with the ability to select the enrollment authority that will define their sample from xLab’s UI.
In order to align xLab with Growthbook’s terminology, and support experiment owner’s transition to the latter third party experimentation tool, we’ll base our design off of the “Identifier types” field available in that external platform:
Quick overview of requirements:
| Placement | Default state & Validation | The field is not modifiable after activation |
  |  |   |
| The “User identifier type” field is the first field included in the “Traffic and variations” section (fka “Sampling”). It replaces the “Sample unit” field in the experiment form and read-only view: | No option in this field is selected by default. If users try to submit without having made a selection, validation should be triggered | Users won't be able to edit their selected user identifier type once their experiment has been activated: (1) We'll mention this in the activation confirmation dialog. New message: "Activating the A/B test {A/B test name} will enable data collection on {start date}, provided instrumentation code has been deployed to production. After activation, the experiment's name, traffic, and variations configuration will no longer be editable". – (2) The User identifier type options will display a disabled state when users try to edit an active experiment. |
Please find detailed specs in this Figma file.
Acceptance criteria
- The new ‘User identifier type’ field replaces the current 'Sample unit' field on xLab's experiment form
- xLab users can employ the required and non-modifiable ‘User identifier type’ field to define the composition of their experiment traffic
