$1.4M net-new spend unlocked
measurement rebuilt, CFO argument settled
Results from specific engagements. Every account starts with an audit because outcomes depend on your product, margins, and creative.
01The symptom
Meta's reported ROAS had drifted 2.8× away from the brand's actual revenue. Every monthly board meeting ended in an argument about which number was real. The CFO was pushing to cut paid spend in half.
02The diagnosis
Both sides of the board argument were defending estimates. The previous setup ran pixel-only tracking that bled 30–40% of conversion events to Safari and ad blockers, reported from platform-attributed ROAS, and had never once measured Meta's incrementality directly.
03What we did
- Deployed server-side CAPI alongside the pixel, with dedupe logic on Shopify orders
- Built a blended attribution view in the warehouse (Meta, Google, organic, referral), reconciled to actual Shopify revenue nightly
- Ran an 8-week geo-holdout test to measure Meta incrementality directly, not from platform data
“Meta said 4.6. Our books said 1.6. Every board meeting ended in the same argument until the geo-holdout settled it. Our CFO signs off on paid spend now without a fight.”
04The full story
After iOS 14.5, Meta reported a 4.6 ROAS on this account. Measured directly against Shopify revenue, the real figure was closer to 1.6. That 2.8× gap amounted to roughly $400k a month in phantom contribution the CFO could not reconcile. Every board meeting ended in the same argument about which number was real, and the CFO was pushing to cut paid spend in half.
Nobody in that argument had good data. The account ran pixel-only tracking, which was bleeding 30–40% of conversion events to Safari privacy changes and ad blockers. It reported from platform-attributed ROAS, a number Meta grades on its own homework. And nobody had ever run an incrementality test. Both sides were defending estimates. The Head of Growth was arguing from an inflated number and the CFO was arguing from a blind one.
The first fix was plumbing. We deployed server-side CAPI alongside the pixel, with deduplication against Shopify order IDs so no event counted twice. Attribution coverage climbed from 61% to 94% inside ten days. Reported numbers moved closer to reality, not further from it.
The second fix was measurement hygiene. We built a warehouse view reconciling every paid-media dollar against actual Shopify revenue, not platform-reported revenue, with 7-day and 28-day lookbacks refreshed nightly across Meta, Google, organic, and referral. The gap between Meta’s number and the books shrank from 2.8× to 0.5×, inside the noise band of normal attribution drift.
Then we ran the test that settled the argument. An 8-week geo-holdout across six US metros measured Meta’s incremental contribution directly, with no platform data anywhere in the calculation. The answer: Meta was driving about 72% of what it claimed. Worse than the platform said. Better than the CFO feared. Either way, a number both sides could sign.
An honest number changed the budget conversation. The brand unlocked $1.4M in net-new annual paid spend that had been frozen in the dispute, and the monthly board meeting stopped relitigating attribution. Ten weeks, start to finish.