Six ad platforms, no deduplication — attributed revenue was 440% of actual

The situation

A K-Beauty brand selling globally across the US, UK, and Asia was running ads on Meta, Google, TikTok, Snap, Pinterest, and Naver simultaneously. Their CMO was preparing a board report and pulled revenue attribution from all 6 platforms. Total attributed revenue: $4.2M. Shopify actual revenue: $955K.

The problem

Each ad platform uses last-touch attribution and claims credit for any conversion that happens within its attribution window. A customer sees a Meta ad Monday, a Google ad Wednesday, a TikTok ad Friday, and buys Saturday. All three platforms claim 100% credit. With 6 platforms, 440% attribution is mathematically expected if the same purchase paths are in all attribution windows simultaneously.

What we did

1. Built a multi-touch attribution model in BigQuery using Shopify order data as ground truth
2. Mapped each order to all ad interactions within 7-day click / 1-day view window using platform-exported data
3. Applied data-driven attribution: distributed credit across platforms based on incremental lift model
4. Built Looker Studio report showing each platform's attributed revenue vs incremental revenue contribution
5. Identified Pinterest and Snap as zero incremental contributors — paused both, reallocated $18K/month

Why it worked

Self-reported attribution from ad platforms is universally optimistic — every platform claims full credit for every conversion in its window. 440% inflation is not unusual for a 6-platform stack. The only accurate picture comes from Shopify/Stripe revenue as ground truth, combined with incrementality testing to determine which platforms are actually driving purchases vs. just appearing in the journey.