+85% month over month without giving back the margin
An established Egyptian store stuck on a revenue plateau reached EGP 2.45M in monthly sales across 3,475 orders, an 85% month-over-month increase held at profit.
An established store that had stopped growing. Spend was rising month after month and revenue was not following it up, which is the specific shape of a problem that looks like a budget issue and almost never is. The owner's read was that the market had saturated. The account said otherwise.
- Multiple ad sets were bidding into the same audience, so the account was paying a premium to compete with itself at every auction.
- Creative was being judged on gut feel after two or three days, with no threshold written down, so winners were being killed early and losers were being kept out of hope.
- Retargeting was absorbing budget that prospecting needed, inflating reported ROAS while the actual new-customer count stayed flat.
Rebuilt the structure
Campaigns split into a clear top, middle and bottom of funnel with the overlap removed, so each stage was buying its own audience rather than re-buying the last stage's. Consolidating the fragmented ad sets also got the account back out of a permanent learning phase, which had been quietly taxing every test it ran.
Made the testing log binding
Every creative now shipped with a pass and fail threshold written before launch and a date to judge it on. It removes the argument from the decision: a creative that misses its number gets killed whether or not anyone liked it, and one that hits gets budget whether or not it was anyone's favourite.
Stepped budget instead of jumping it
Increases of 20 to 30 percent a week on stable winners only, measured against net profit rather than reported ROAS. Scaling faster than that resets learning and costs more in lost efficiency than the extra reach is worth.
Monthly sales reached EGP 2.45M across 3,475 orders, an 85% jump month over month, with margin intact rather than bought. The plateau had not been the market. It had been the account competing with itself while nobody was keeping score.
The first three weeks looked worse before they looked better. Cutting overlapping ad sets pulls reported ROAS down while the account re-learns, and that is a genuinely uncomfortable conversation to have with an owner watching a dashboard. It is also why the dashboard exists: the fall was visible, expected, and explained in advance.
