A major Saudi retail brand, bleeding.
A year earlier, online and total monthly sales had been far higher. Within six months online sales fell by roughly a third and total sales by about as much, at ROAS 5. No UGC, no influencers, and no time to make any. Every unit of spend had to come back.
The audit found the problem on all four platforms, not one.
Meta and Snapchat ran one repeating scenario: an offer exists, campaigns work; the offer ends, everything collapses. No audience segmentation, everyone saw the same ad. TikTok had not a single video while competitors were doing millions there. Google had Pmax, Smart Shopping and Search all bidding on the same terms, so the account competed against itself.
The worst finding: primary conversion set on every event, so the account optimised toward add-to-cart while believing it optimised toward purchase. Tracking carried a large delay and tampered tags. Every decision was sound reasoning applied to false numbers.
The rule we worked to: if the foundation is broken, tear it down and build a new one.
We started with the numbers, not the ads. Tag Manager helped a little but accuracy stalled at 70%; Magento Tools connected every platform accurately and the delay effectively disappeared.
Then Google from zero: clean Search with one ad group for categories and one for brand, Standard Shopping for core products, Feed-Only Pmax for speed, then Pmax split by strategy, one campaign for the three available videos, one for visitors of direct competitors' sites, one using all assets. Each campaign with a single job. Inside a month ROAS held and volume climbed, and we raised budgets 25% every four days.
On Meta, budgets were set against break-even ROAS and AOV rather than instinct, split into acquisition, re-engagement and retargeting. The surprise: despite the account's age, retargeting had never been built properly. Once it ran, the numbers moved.
On Snapchat the bidding strategy was the lever: testing 50% to 150% of target CPP, target cost won. On TikTok, three branch videos, value optimisation, and audiences split by high spending power and iOS.
Two experiments moved overall ROAS more than anything else. First, format-based campaigns: instead of splitting by volume or leaving placements automatic, we split by format, one campaign for 1:1 square only and one for 4:5 and 9:16 vertical only, with no mixing. Second came from the data rather than from opinion: a large share of orders used Tamara. Saudi customers like instalments, so a dedicated campaign explaining the limit and the steps became one of the highest-ROAS campaigns in the account.
Results: online revenue up 139%. ROAS from 6 to 11.8. Spend from 600,000 down to 565,721 (−8.4%). Orders from 2,700 to 3,392. CPA down 52% to 166. Combined online and offline sales more than doubled.
Spend fell and revenue doubled, not because of a magic button, but because the account stopped fighting itself.
The lesson
If the foundation is broken, rebuild it
