EMQ 2.4 to 9.1: Fixing Meta Attribution When 71% of Your Customers Are on iOS

The situation

A US baby products brand was spending $28K/month on Meta Ads. Their customer base skewed heavily toward parents — and 71% of their customers were on iOS. Their Meta Event Match Quality was 2.4 — one of the worst scores possible. The algorithm had almost no reliable data to work with.

The problem

An EMQ of 2.4 means Meta can barely match browser events to real user accounts. On iOS, without user consent for tracking, Meta's pixel receives an anonymized event with no identifying information. Without server-side first-party data enrichment, every iOS purchase arrives at Meta as a completely anonymous signal — impossible to attribute to a specific ad, campaign, or audience. At this score, smart bidding is essentially guessing, lookalike audiences are built on unreliable profiles, and ROAS reporting has almost no connection to actual business outcomes.

What we did

1. Integrated Klaviyo as a first-party data source — on every Shopify order, Klaviyo automatically provides the customer's email, phone, and postal code in hashed format
2. Rebuilt CAPI to populate all available PII parameters on every purchase event: email, phone, first name, last name, city, state, ZIP, country
3. Added Snapchat CAPI simultaneously using the same GTM Server-Side infrastructure — both channels fixed in the same engagement
4. Ran a 30-day post-implementation review comparing EMQ scores and purchase event volumes week over week to confirm sustained improvement

Why it worked

An EMQ of 9.1 is exceptional — it means Meta can confidently identify and attribute nearly every purchase to a specific user account, even on iOS. The key is first-party data: when Saifur enriches every CAPI event with hashed email and phone (pulled from Klaviyo's order data), Meta can match that event to its user database regardless of whether the customer had tracking consent enabled on their device. This turns an iOS-heavy audience from a tracking liability into a strength.