Meta's algorithm is exactly as good as the data you feed it. Since iOS 14.5, a pixel alone misses a significant share of conversions: ad blockers, Safari, consent refusals. The result: you are paying to optimize an algorithm that sees blurry. Here is how to check if that is you.
What the pixel no longer sees
The pixel is JavaScript in the browser. Anything that blocks the browser blocks it: ad blockers, Safari and Firefox tracking protection, consent banner refusals, iOS apps with App Tracking Transparency declined. Depending on your audience, that is easily 20 to 40% of conversions invisible to the pixel.
“A pixel alone misses 20-40% of conversions: CAPI is half the signal, not a bonus.”
CAPI is not optional, it is half the signal
The Conversions API sends the same events from your server, where nothing gets blocked. With hashed customer data (email, phone), Meta reconciles the purchase with the ad click. The quality of that reconciliation is measured in Events Manager with the EMQ score (Event Match Quality): below 6/10, you are leaving performance on the table.
Deduplication: where everyone gets it wrong
Pixel + CAPI = every purchase sent twice. Meta deduplicates only if both events share the same event_id. Without it, your purchases count double, your platform ROAS inflates artificially, and you scale campaigns that do not actually perform.
The test: Events Manager, Purchase event, look at the deduplication processing line. If you see two sources without merging, it is broken.
The 4 signs your tracking lies
1. Platform ROAS diverges more than 30% from your MER (total revenue / total ad spend) with no explanation.
2. Meta reports more purchases than Shopify has orders over the same period.
3. Your EMQ score is under 6/10 on the Purchase event.
4. Your conversions dropped sharply on an iOS update and never really recovered.
The 30-minute fix
Open Events Manager and Test Events. Make a test purchase. Check: does the event arrive from both browser AND server? Do both carry the same event_id? Is Purchase EMQ at 6+? Then compare 7 days of platform conversions against your backend. If the gap exceeds 15%, the problem is your tracking, not your creatives.