ROI and analytics problems that come up again and again, worked through with the math shown.
Illustrative scenario. The company and all figures are hypothetical and show how we’d approach the problem. This isn’t a client result, and your numbers will differ.
A subscription app sees a healthy 1.90 reported ROAS. Measured against sales that would have happened anyway, the real figure is 1.30, and that changes where the budget should go.
1.90 Reported ROAS (hypothetical)
A $5 cost per install on brand terms looks like the best deal in the account. A holdout suggests the true cost is $12.50, and moving part of that budget buys 25% more incremental installs.
$5.00 Reported brand cost per install (hypothetical)
When one ad takes most of the budget, broad targeting has little to work with. More distinct concepts take this hypothetical account from $133 to about $106 per paid subscriber.
$40 to $32 Cost per trial (hypothetical)
Moving App campaigns from installs to a downstream in-app event raises the cost per install from $2 to $5, yet this hypothetical app ends up with 30% more subscribers from the same budget.
$2.00 to $5.00 Cost per install (hypothetical)
Three tools report three different counts of the same paid conversions. A shared definition and a few fixes narrow the gap from 15% to about 3% and cut weekly reporting from two days to two hours.
±15% Gap to billing data, before (hypothetical)
A mix model credits Meta with a 1.8 return, but a four-week geo holdout measures 1.2. Calibrating the model changes the budget split and adds an estimated 3% incremental revenue.
1.8 Model's Meta return, before (hypothetical)