Meta was reporting 2.23 times the revenue the store’s own records credited it with, and nobody could say which number to spend against. We built the store its own order ledger, ran 155 days of spend through it, and measured what each platform’s next dollar buys. Meta prospecting came back at 1.92x. Two campaigns Growth Mate had built itself came back at zero.
Key results.
July 2026 is the strongest documented month, and the first read on two attribution lenses.
Challenge
Meta carries almost the entire budget. In July 2026 it took $497,468.70, 94.7% of the money, and reported $961,899.93 back at 1.93 times spend. The store’s own order records tie that same spend to $431,750.72 across 6,220 orders, a 0.87 return, so Meta reports 2.23 times what the store credits it. Both are correctly calculated, and neither answers the question that decides a budget: what would have happened without the spend.
The store had no own-data attribution to check either number against. A 98-day audit of 61,957 orders found the largest leak was a self-referral carrying no UTMs, worth $2,502 a day and ongoing, roughly 6% of daily revenue, plus $598 a day of coupon-extension interception.
The account had also been damaged from the inside, some of it by us: Google Ads fell to $139 a day in early July 2026, self-inflicted per the change history. In March 2026 Meta’s Health and Wellness enforcement hit the cortisol health-claim copy and collapsed spend 82% and purchases 93%, against a product carrying more than 50% of revenue.
Solution
Step #1 Build the store its own ledger.
We replayed every Shopify order through the admin app proxy into a UTM order ledger, so the store could read its own last-click attribution. It grew from 17.5k orders in July to 95,750 by September.
Step #2 Measure the spend instead of assuming a coefficient.
On Simon’s instruction to stop assuming a 0.7 PMax coefficient and measure the real ones, we ran a 155-day daily panel from 1 April to 4 September 2026 across 95,750 orders: HAC and Newey-West regressions, plus event studies on every historic budget move, read off the Meta ad-account activity log that nobody had been using as a change history. Meta prospecting came back at 1.92x, 95% CI 1.78 to 2.05, t of 28, with five clean dose events pooling to 2.01x, above the 1.45x break-even. Google PMax came back at -0.26, interval -0.76 to 0.24, and Demand Gen at -0.13. Both are zero, and both were campaigns we had built.
Step #3 Act on the measurement, including against ourselves.
Google was rebuilt in a single day on 8 July, 1,156 of the month’s 1,310 changes landing then, nine campaigns opened and eleven turned off. Spend moved out of one narrow Performance Max campaign, which fell from $9,650.56 to $1,204.40, into a Top Sellers structure that took $14,375.80 and returned 2.95 on the store’s own records across 593 orders. Inside the same month we turned off three Meta ads carrying $29,389.25 at 1.34, the largest of them the account’s second-biggest ad at $23,049.96 and 1.38, and eleven of our own Google test campaigns at 0.45. That is 6.3% of the month’s budget stopped while the month was still running, and on 7 September 2026 a seven-line reallocation went live at constant total spend, verified by read-back.
Results
Book a discovery call about your business goals
- ✓Non-binding introductory meeting
- ✓Explore your growth potentials
- ✓Talk to our experts