How the numbers actually happened.

A new brand launching cold into the Egyptian market. No pixel history, no creative library, no purchase events for the algorithm to learn from, and no evidence the offer would hold up once it met traffic that had never heard of it. Everything a normal optimisation playbook depends on was missing, so the first job was not scaling. It was building something worth scaling.

• The pixel fired on page load rather than on the confirmed order, which would have taught the algorithm to optimise for people who browse.
• The product page led with the discount instead of the reason to buy, so the offer had no argument behind it if the discount ever came off.
• There was no creative library at all, which meant the first week would be a hypothesis test, not a launch.

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.

A high-ticket brand where cheap traffic was actively harmful. At this price point a low cost per click is usually a warning: it means the ad is reaching people who will never buy, and every one of them teaches the algorithm to find more like them. The brief was qualified buyers at the price point, not volume.

• Creative was priced-led, which pulled in discount-seekers and dragged the average order value down every time it scaled.
• The audience signal was built on reach rather than intent or value, so the account optimised toward the cheapest available human.
• The owner had no per-product profit view, so a product with a good ROAS and a bad margin looked identical to one with both.

A brand doing real volume, with an owner reasonably pleased: the ads manager showed 5.48 ROAS and the blended figure across the whole business looked close to 10x. The plan was to raise budget. The problem with raising budget against a number you have not audited is that scaling does not create margin, it multiplies whatever the account is already doing — so if the number is wrong, the mistake gets bigger at exactly the rate the spend does.

• The account was reporting on a 7-day-click-and-1-day-view window while the owner compared it against same-day revenue, so a week of purchases was being credited to a single day of spend.
• Not one ad carried a UTM, which meant Shopify could not attribute a single order back to the campaign that bought it. Every platform number was unfalsifiable by design.
• COD and Fawry orders were both counted as revenue the moment checkout completed, before either had settled — and in this market a meaningful share of them never does.

Good product, decent creative, clean traffic, and almost nothing coming out the other end. The instinct in that position is always to make more ads, and it is usually the wrong instinct: when the funnel is full at the top and empty at the bottom, the ads are doing their job and something after the click is not. A brand cannot outrun its own price list with a better hook.

• Five direct competitors selling a comparable satin set had all converged around EGP 1,150, which is what a market price looks like once a category matures. This brand was at EGP 1,300 — around 13% above, and visible on the first product page any shopper compared.
• The real cost per order sat near EGP 300 against an average order value of about EGP 1,750, so the account had margin to work with. The blockage was never efficiency.
• Everything was sold as a single item, which left no way to raise order value without raising the price of the thing the buyer was already hesitating over.

A large catalogue — 618 live products — being advertised without an agreed definition of a good result. That sounds like a bookkeeping detail and it is the whole game: without a break-even, every ROAS is just a number to have opinions about, and an account can be scaled confidently for months in a direction that quietly loses money on delivered orders.

• The true break-even was 1.77 ROAS, and 2.11 once the cancellation rate was included. Campaigns clearing 1.9 had been treated as winners and were losing money on every delivered order.
• Realised discounting ran 16.2% deeper than the discount the store thought it was giving, through stacked codes and legacy promotions nobody had switched off.
• Benchmarked against the closest comparable Egyptian silver retailer, the catalogue was priced around 27% higher — which is a viable position, but not an accidental one, and nothing in the ads was doing the work to justify it.