The meeting every marketing team has had
Meta’s ads manager says the campaign drove 120 purchases. GA4 says 70. Your shop backend says 95. Someone asks “so which number is right?” - and the honest answer is: they’re all right, about different questions. Nobody is lying. The tools measure different things, on purpose.
Once you know the specific reasons they diverge, the mismatch stops being alarming and starts being useful. Here they are.
The five reasons the numbers differ
1. Attribution windows. Meta claims a conversion if someone clicked (or even just viewed) your ad within its window - by default, 7-day click and 1-day view. GA4 doesn’t count view-through conversions from Meta at all. A person who saw your ad Tuesday and bought via Google search Friday counts for Meta and for Google - and each platform sincerely believes it.
2. Who gets the credit. Meta only sees Meta. It credits itself for every conversion inside its window, regardless of what other channels did. GA4 sees all your channels and splits credit across them. Platform numbers are self-graded homework; GA4 is the cross-channel referee - with its own model and blind spots.
3. The conversion date. Meta books the conversion on the day of the click; GA4 books it on the day of the purchase. On any given day or week the two are counting different sets of people. Comparing short windows makes the gap look wild; over a month it usually narrows.
4. What each tool can see. Ad blockers and consent choices hit the two tools differently. Meta backfills part of the gap with modeled conversions; GA4 (with Consent Mode) models differently, or not at all depending on your setup. The gap between them is partly a measurement-technology gap, not a marketing one.
5. Definitions and duplicates. “A purchase” sounds unambiguous until one tool counts a refunded order, the other dedupes a double-fired event, and your backend counts neither. If events aren’t defined once and reused everywhere, every tool invents its own version.
Which number to use for which decision
- Optimizing inside Meta: use Meta’s number. It’s what the algorithm learns from, and feeding it better data (that’s what CAPI is for) is the highest-leverage fix.
- Comparing channels against each other: use GA4 - it’s the only one of the two that sees everything - but read it knowing it under-counts view-driven and dark-social effects.
- Reporting revenue to anyone: use your shop or CRM. It’s the only source that knows what a real order is. Marketing numbers should reconcile toward it, never replace it.
The real fix for a small team
The gap itself isn’t the problem - an unexplained gap is. A healthy setup states its expected differences (“Meta reports roughly 25% above GA4 because of view-through”) and alarms when reality drifts from that. That takes clean event definitions, GA4 configured to match how you actually make money, and a regular reconciliation against your backend.
That reconciliation work is a standing part of what we run for teams - GA4, set up so it agrees with the rest of your numbers. Or do it yourself: pick one week, trace ten real orders through all three systems, and you’ll learn more about your tracking than any dashboard will tell you.