Skip to content

BasicBucketerExperiment

One experiment’s arms, and optionally how much of the population sees it.

optional exposure?: number

Fraction of subjects enrolled at all, 0..1. Default 1 (everyone).

A subject outside it is reported as NOT ENROLLED — the experiment is simply absent from the assignment map, which is distinct from being assigned to a control arm. That distinction is the whole point of a holdout: control is a measured arm, unenrolled is not in the experiment.


variants: readonly string[] | Readonly<Record<string, number>>

Either an equal split across the listed variant handles, or explicit relative weights. Weights need not sum to anything in particular — they are normalized — so { control: 1, variant_b: 3 } is a 25/75 split.

A variant weighted 0 (or negative) is excluded rather than made unreachable-but-present, so removing an arm and zeroing it behave the same.