The first week of November: 9,008 orders, ROAS 9.35, sales up 150%.
But the first problem solved in this account wasn't creative or budget. It was that platform-reported numbers were far higher than actual sales. Running everywhere means every platform tells you it produced the sale, and adding their figures together gives you more revenue than the bank has.
So we focused on UTMs per platform and tracked through Mixpanel and Analytics: who actually delivers, and who is simply claiming credit. As each platform scaled, we watched what genuinely increased sales versus what made no difference to MER. Testing showed TikTok and Google as the strongest, with Snapchat alongside, and budget was allocated accordingly.
On Google, Pmax was spending most of the budget on brand terms, paying to reach people already searching for you by name. A negative keyword for the brand name inside Pmax fixed it, and something odd followed: Pmax ROAS dropped sharply while Search rose significantly, and we ended up spending less for the same return. With competitors appearing on the brand name, an impression share campaign with a higher-than-normal target CPC secured the top result on our own name.
On Snapchat and TikTok, testing ran on auto bid with two or three variations per campaign using different hooks, which raised efficiency and slowed creative burnout. Winners moved to target cost, max bid and target ROAS on TikTok, raised at 2x, 5x and 10x with every action monitored. Once the failures from testing were closed, an evergreen set remained across all platforms.
In scaling, timing mattered most. Allocation followed the platforms genuinely delivering, tracked hour by hour through UTMs, and campaigns were not heavily punished at day's end, because the early morning hours produce 20 to 25% of daily sales. Scaling starts with the new day.
Constants in the account: catalogue on Meta and Snap, Shopping on Google for prospecting with retargeting through Pmax, and retargeting via WhatsApp. Whitelisted testing ran videos from the influencer's account and ours simultaneously, with three variations pulled from every winner.
Behind all of it sat material produced to a required direction and a very large operation.
With attribution right, you raise budget on the platform that delivers. With attribution wrong, you raise budget on the platform that talks the loudest.
The lesson
Attribution before scale
