These experiments are run all the time even before the internet. I remember reading about how for broad brand advertising they used to segment by city. Half the cities got an ad for Coca Cola and half didn't. Then you compare sales of Coca Cola. No one will publicly publish numbers because it's a mix of sensitive sales data and competitive advantage (ie: otherwise your competitors don't have to burn money running their own tests). Also, smaller shops turn their online advertising campaigns on and off all the time to test impact.