A randomised experiment in which some eligible people are withheld from the activity, and the difference in outcome between the exposed and the withheld is the effect.
Also called Randomised controlled trial, A/B test
Compose from here— 6
Holdout experiment can stand in 6 roles across the patterns this model declares. Pick one and the answer is drawn here, from the published graph.
- board → holds KPIs, Metrics · the question
- board → holds Metrics, Time Windows, Geographies, Audience Segments, Units · the shape
- board → holds Attribution Models · the question
- board → holds Metrics, Parameters · the shape
- board → holds Parameters · the question
- board → holds Parameters · the question
Related— 3
- calibrates →Attribution lookback windowParameters
- calibrates →Incrementality factorParameters
- grounds on →MeasureEntities
Pointed at by— 5
- parameterized by →Minimum detectable effectParameters
- parameterized by →Statistical powerParameters
- parameterized by →Significance levelParameters
- parameterized by →Holdout shareParameters
- establishes →Reactivation RateMetrics