The first eighteen months after Apple's App Tracking Transparency rollout were chaos. The next eighteen months were the long, quiet rebuild. We are now in a stable place — not because attribution is "fixed," but because the operators who survived have settled on a working stack. This is the stack we run inside active fashion D2C accounts in 2026.

What is dead

Some things that used to work do not. Pretending otherwise wastes spend and produces bad strategic decisions.

  • Pixel-only attribution. Browser tracking lost 30–40% of conversion signal post-iOS 14.5 and never came back.
  • 7-day-click attribution windows. Use 7-day-click + 1-day-view as your default — most fashion purchases sit in the consideration window.
  • Lookalike audiences below 1%. The algorithm consolidates them now. Set your floors at 1–5% and stop fragmenting your audience pool.
  • "Best practice" UTM hygiene without server-side reinforcement. Marketers are still relying on UTMs as if they survive the iOS sandbox. They do not.

The 2026 signal stack

Every account we manage runs four overlapping layers:

Layer 1 — Conversions API (CAPI)

Server-side event firing direct from the e-commerce platform. Implemented natively if the store is on Shopify Plus, via Stape or CAPI Gateway for everything else. Deduplication keys are non-negotiable. This is the foundation — everything else builds on top.

Layer 2 — First-party CRM enrichment

Hashed email and phone on every event, pulled from the customer record at purchase. This restores match rates that used to come from pixel cookies. Without it, Meta is matching on probabilistic signals — which is exactly what the post-iOS world is supposed to eliminate.

Layer 3 — Blended attribution model

In-platform ROAS shows you Meta's view of the world. Blended ROAS (total revenue ÷ total ad spend across channels) shows you the actual business. We run both on every account and look for the gap. When in-platform ROAS is 3× higher than blended, you are double-counting. When blended is 30% higher than in-platform, Meta is under-reporting and you have room to scale.

Layer 4 — Triangulation with third-party tooling

We use a triangulation layer (Northbeam, Triple Whale, or a custom warehouse-based attribution model) to cross-reference Meta's reporting against the broader marketing mix. This is not for tactical decisions — those still come from Meta's view. It is for monthly strategic decisions about channel mix and budget allocation.

What is actually working

Despite the attribution noise, the things that move fashion accounts forward have not changed. Strong creative tested rigorously. Audience architecture built around intent. A relentless focus on retire-and-replace cadence. Attribution is the foundation under all of that — when it is solid, everything else compounds. When it is broken, every optimization sits on quicksand.

“Match rate is the most important number nobody is talking about. Below 85% on purchase events, no other optimization matters until that is fixed.”