Similar to the experiment targeting, when variant targeting is activated, users only see a variant if their targeting rules are met.
If targeting is only defined for some variants of an experiment, the evaluation follows the following scheme:
1. In the first step, all variants are evaluated with existing targeting. Among all variants whose targeting does not apply, one is randomly drawn.
2. If the first step has not determined a variant, a random one is determined from all variants with empty targeting.
Variant targeting and traffic distribution cannot be active simultaneously.