Clean Measurement: Meta CAPI + GA4 Without Double Counting

Running Meta Conversions API (CAPI) alongside GA4 sounds simple. But if you're not careful, you'll double-count every conversion. We've set up this stack for dozens of clients. Here's how we do it right. The Problem When you run both server-side (CAPI) and client-side (Meta Pixel) tracking, the same event can fire twice: - Client-side: Browser fires pixel event - Server-side: Server fires CAPI event If both events count, you're seeing double the conversions you actually have. That breaks attribution, inflates ROAS, and makes optimization impossible. The Solution: Event IDs The fix is event deduplication using event_id. Both Meta Pixel and CAPI support an event_id parameter. If you send the same event_id from both sources, Meta deduplicates automatically. Here's how we implement it: 1. Generate a unique event_id for each event (UUID v4) 2. Send the same event_id from both client and server 3. Meta automatically deduplicates on their end The event_id needs to be: - Unique per event (not per user or session) - Consistent across client and server - Stored so we can reference it later We generate event_ids client-side, then pass them to the server via webhook or API call. The server uses the same event_id when sending to CAPI. Server vs Browser: When to Use What Not every event should fire from both places. Here's our rule: Server-side (CAPI only): - Purchase events (most reliable from server) - Lead form submissions (server is source of truth) - High-value events (we want guaranteed delivery) Client-side (Pixel only): - Page views (browser-only context) - Video views (browser-only context) - Scroll depth (browser-only context) Both (with deduplication): - Add to cart (we want fast feedback + reliable backup) - Initiate checkout (same reason) - View content (browser for speed, server for reliability) The key is understanding which events need redundancy and which don't. QA Checklist Before we go live with any CAPI + GA4 setup, we run through this checklist: 1. Event IDs: Are they unique? Are they consistent across client/server? 2. Event names: Do they match between Pixel and CAPI? 3. Parameters: Are all required parameters present? Do they match? 4. Test events: Can we see test events in Meta Events Manager? 5. Deduplication: Are events being deduplicated? (Check Events Manager) 6. GA4: Are events firing in GA4? Are they not duplicating? 7. Attribution: Are attribution reports showing correct data? We test this in a staging environment first, then validate in production with test purchases. Common Pitfalls We've seen these mistakes over and over: - Missing event_id: Events fire but don't deduplicate - Inconsistent parameters: Events deduplicate but data doesn't match - Timing issues: Client fires before server, or vice versa - Missing test mode: Can't validate before going live Our solution: a unified tracking library that handles all of this automatically. Client and server use the same code, same event structure, same validation. The Bottom Line Clean measurement requires discipline: proper event IDs, smart server/browser split, thorough QA, and avoiding common pitfalls. When done right, you get accurate attribution, reliable data, and the ability to optimize with confidence. When done wrong, you get double-counted conversions and broken reports. The difference is in the details.