Incrementality analysis
Estimate effects by comparing exposed and control groups, or periods before and after advertising.
Keep the mappings, filters and model options in your project.
Estimate differences against a comparison group or prior trend. Whether they represent a causal ad effect depends on the design.
Control groups compare exposed and unexposed users concurrently and provide the strongest design. Launch and shutdown compare before/after; a control group helps separate seasonality and trend.
Compares a holdout group (ads randomly blocked) vs an exposed group over the same period. Random assignment gives the highest causal confidence.