Growth Engineering Playbook · Wave 4
The channel reported the sale. It didn't cause it.
Every platform reports the conversions it touched — but touching a sale isn't causing it. The only honest way to know what your spend actually added is a holdout: withhold the campaign from a random control group and measure the gap. This turns seven fictional Northstar Outfitters holdout tests into true incremental lift — a two-proportion z-test with confidence intervals, and reported-vs-incremental ROAS — so demand-harvesting channels stop taking credit for sales that were coming anyway. Deterministic, client-side.
Loading experiments…
What the spend actually caused
Hatched = baseline the control group converted anyway. Solid green = true incremental lift.
Is the lift even real?
One experiment's holdout, tested.
Reported vs incremental, by channel
Reported ROAS is what the dashboard shows. Incremental ROAS is what the holdout proves. The verdict is the two-proportion z-test at your confidence level.