> You can simply control by making only one single change segregated to a control and treatment geo/group before and after, measure delta. Because nothing else

> You can simply control by making only one single change segregated to a control and treatment geo/group before and after, measure delta. Because nothing else is going on in the world besides your ad campaign, right? Like you I find the comments about the ad industry here on HN largely ignorant, but I also think most advertisers and agencies are dumb as rocks, or at best high on their own supply. Like, a piece of analysis I saw recently had some claimed weirdass causation along the lines of, X days is the optimum period between first indication of intent in a product and showing an ad for that product, if the user didn't convert immediately. They had run an experiment like you suggest with varying intervals for audiences. But, the way better correlation was with stimulus payouts. That wasn't in their model, because their model only had the inputs they controlled, before and after, and measured delta. Everyone is selling snake oil because every advertiser is demanding snake oil. It's so fucked up.