No currency codes on Meta revenue — the best market looked like the worst

The situation

A fashion brand based in Dubai was selling across 5 markets: UAE (AED), UK (GBP), Germany (EUR), USA (USD), and Australia (AUD). They were spending $45K/month across these markets on Meta and Google. Their head of performance had been trying for 6 months to identify their most efficient market — but comparing a 3.2x ROAS in GBP to one in AED was meaningless, because the numbers were in different currencies.

The problem

Their Shopify Markets store passed purchase revenue to Meta and Google as the local currency amount — with no currency code. Meta and Google assumed it was in their billing currency (USD). A £180 UK purchase was processed as $180. A AED 650 UAE purchase was treated as $650 — nearly 2.5× its real dollar value. Across 5 markets, every single ROAS figure in every campaign was wrong by up to 150%. Budget allocation was built on currency fiction.

What we did

1. Audited all 5 Shopify Markets stores — confirmed purchase events were passing local currency amounts without currency codes
2. Built a server-side currency normalization layer in the GTM Server-Side container: a custom variable fetches daily exchange rates and converts all revenue to USD before passing to any ad platform
3. Added the currency: "USD" parameter to all Meta and Google event payloads
4. Built a unified Looker Studio dashboard showing all 5 markets with USD-normalized ROAS — direct market comparison now possible for the first time
5. Retroactively calculated correct ROAS for the prior 6 months using accurate exchange rates to establish a true baseline

Why it worked

Multi-currency is completely invisible without looking for it — every platform accepts whatever revenue number you send without validating the currency. The UAE market had been systematically underfunded because AED values are numerically high relative to USD, making UAE campaigns appear to overclaim. In reality, UAE had the highest true ROAS of any market. An entire market was underfunded because of a number-format error.