From 12% to 71% Match Rate: Why TikTok Attribution Was Broken

The situation

A Canadian pet products brand was spending $18K/month on TikTok. Their pixel match rate had been stuck at 12% since launch. Creative teams were making decisions about which videos to scale and which to cut — but the data they were using was almost completely unreliable. One account manager described it as "flying with one eye closed."

The problem

TikTok's browser pixel relies on being able to identify the user in the browser — but on iOS (with tracking transparency), desktop browsers with privacy extensions, and Safari's ITP, this identification fails. At a 12% match rate, TikTok could only confidently attribute 1 in 8 purchases to the correct campaign. The other 7 were being distributed by guesswork. The algorithm had no reliable signal to learn from, meaning even creatives that were genuinely working couldn't be identified as such.

What we did

1. Set up a GTM Server-Side container on a custom subdomain — this allows events to be sent from the server instead of the browser
2. Configured the TikTok Events API to route all purchase, add-to-cart, and page view events server-to-server, completely bypassing browser limitations
3. Added SHA-256 hashed email and phone number enrichment to every event payload — TikTok uses these to match events to logged-in TikTok accounts
4. Implemented pixel-to-API deduplication using unique event_id values to prevent any event being counted twice

Why it worked

TikTok's Events API sends data server-to-server — it doesn't rely on the user's browser at all, so iOS privacy settings and ad blockers have no effect. When you combine that with hashed PII data (email and phone), TikTok can match events to logged-in TikTok users across any device. A 71% match rate means the algorithm can now learn from the majority of real customer behavior — not a random 12% sample.