How the numbers actually happened.
Six accounts, written the way I would explain them to another media buyer: what the account looked like before, what I found when I opened it, what I changed, and what the change did. Brand names are withheld to honour NDAs. The numbers are not edited.
Each one ends with the part most case studies leave out — what was uncomfortable, what I would not repeat, and where the result owes more to the offer than to the advertising. A page of unbroken wins tells you nothing about how somebody works.
How to read every number on this page.
A result quoted without its measurement rules is not a result. So here are the rules these were measured under, in advance of anyone having to ask for them.
The numbers come from inside the ads manager and the store
Sales and order counts are read from the platform and the store back end for the stated window, not from a designed report. The raw screenshots sit on the homepage with dates visible.
Attribution window is stated, not assumed
Meta's default 7-day click / 1-day view window flatters every account it touches. Where a figure is platform-reported it is treated as platform-reported, and the decision that followed it was made against the store's own revenue.
Cash on delivery is counted after cancellations
In Egypt a meaningful share of confirmed orders never gets delivered. A ROAS calculated before cancellations and returns is a number that cannot pay a supplier, so scaling decisions in every account above were made on delivered, net figures.
Brand names are withheld, numbers are not edited
Client work is under NDA, so verticals are named and brands are not. Nothing is rounded up, no window is chosen after the fact to make a figure look better, and the months that did not work are not on this page because they are not case studies — which is worth saying out loud.
