The $17K/Month Tracking Leak Nobody Noticed for a Year

The situation

A US supplements brand was running $45K/month on Meta. Their ROAS had been declining for 3 consecutive months with no apparent cause. Their previous agency had audited the setup twice and both times confirmed the tracking was "working correctly." The client contacted me as a last resort.

The problem

The pixel was firing — but on completely the wrong page. Someone had set up the purchase event on the /cart URL instead of the order confirmation page. This meant every time any visitor added something to their cart, Meta counted it as a completed purchase. Actual purchases — happening on Shopify's /thank-you page — were never being tracked at all. 38% of real purchases were invisible to Meta. The remaining 62% being tracked were people who added to cart but never bought.

What we did

1. Complete GTM container audit — mapped every tag, trigger, and variable, traced each firing rule to its actual URL condition
2. Confirmed the pixel's purchase trigger was firing on /cart — not the Shopify order confirmation page
3. Rebuilt the purchase event on the correct /thank-you URL, then added server-side CAPI via Shopify webhook as a reliable backup layer
4. Set up deduplication, then ran both simultaneously for 7 days comparing against Shopify data before declaring the fix complete

Why it worked

The original implementer had confused the cart page URL with the purchase confirmation URL — an easy mistake that's almost invisible without a systematic audit. The pixel appeared to be "firing" in every diagnostic tool because it was firing — just on the wrong event. This is exactly why I always cross-reference Event Manager data against actual Shopify orders before calling a setup healthy.