A holdout group is a comparison group deliberately withheld from the intervention being studied. In advertising lift measurement, Google compares conversions in a treatment group with conversions among users held back from seeing the ads.[1] This helps investigate what would happen without the campaign rather than treating every purchase after ad contact as additional sales.
In a hypothetical store experiment with a reactivation message, a randomly selected subset of eligible customers could receive no message. Keep other conditions as comparable as possible and measure outcomes over the same follow-up period. Account for different group sizes and customers encountering the same message through another channel. A holdout is a research design, not a guarantee that every observed difference was caused by the intervention.
Content updated:
Sources
Compiled with AI assistance using public sources. General explanation, not individual legal or tax advice. Describing a concept does not mean Zoloo offers it as a feature.