A geo holdout you can run in two weeks

Incrementality testing has a reputation for needing a data team and a quarter of runway. For a $40k-a-month account it needs a spreadsheet and two weeks.

Pick regions that together take 15–20% of your spend and that look like the rest of the country on order value and repeat rate. Match them to a control set on the same measures over the last eight weeks. Turn the test regions off completely — not down, off, because a reduced budget still buys attention and blurs the read.

Run for two full purchase cycles. Then compare the test regions' revenue against the control regions' over the same window, indexed to the pre-period. The gap is what the advertising caused. Divide by what you would have spent and you have an incremental return you can actually put in front of a CFO.

It will almost always be lower than the platform number. That is the point: you now know by how much, and you can set targets against the real figure instead of arguing about the reported one.