Personalizing or A/B testing a section in place
Wrapping a section in a variant container makes it personalized or split-tested — the section becomes the Default arm, unchanged until arms are added.
Wrapping a section turns it into a personalized or split-tested container without changing what a visitor sees: the section becomes that container's Default arm, so the page renders exactly as it did until a second arm is added.
Personalization or an A/B test
Which of the two you're setting up is chosen once, when you wrap the section, and it decides whether an experiment can sit behind it. Wrapping a section as a test does not create an experiment: it gives the container a placeholder experiment id, so the field looks filled in while it points at nothing, and the section keeps showing its Default arm to everyone. The experiment itself is created under Admin → Audience → Experiments and named on the container, with each arm's key matching one of its variants — see Running an A/B experiment. A personalization container never has an experiment at all; there's nothing to find for it under Targeting.
What can't be wrapped
A section that's already inside another container's arm can't be wrapped again — the action is refused outright rather than left to produce a container nested inside a container. A personalized or tested section is always exactly one level deep.
Where arm content actually gets edited
The container's own settings panel is identity only: its mode, its experiment wiring, the list of arms and their traffic split. What an arm actually shows is edited on the canvas itself, in place, exactly the way you'd edit any other section — there's no separate content form for a variant.
Traffic split
Each arm's share of traffic is a fraction of the whole, and the arms on one section have to add up to 1 within a small rounding allowance. That check exists to catch a typo before you save — the number that actually decides who sees which arm at runtime lives on the experiment record, not on this split.