At least once a month I get this question: "Google Ads says 120 sales, GA4 says 87, Meta says 95 — which one is lying?" My answer never changes: none of them. All three count honestly, by their own rule sets; the problem is our assumption that they count the same thing. The assumption looks innocent but its consequences are expensive: shifting budget to the wrong channel, killing a campaign that works, presenting the sum of three platforms to management as "total sales."
Why they will never match
There are four structural reasons. Attribution windows: Google Ads' default counts 30 days post-click, GA4's data-driven attribution distributes credit across touchpoints, and Meta adds view-through conversions within its own window. The same sale appears with three different weights. Counting time: Ads books the conversion to the click date, GA4 to the event date — at month ends this alone creates real gaps. Identity: each platform tracks users with its own cookie/ID system; users who decline consent, switch devices or switch browsers break the chain in different places. Modelling: conversions of unconsented users are modelled differently in each system. Stack all four and a 20–40% gap is normal; expecting agreement misunderstands what these systems are.
The most expensive mistake: summing platforms
"Google 120 + Meta 95 = 215 sales" — that addition counts every sale both platforms claim credit for twice. A customer whose last click was Google but who saw a Meta ad on the way exists in both reports. Your real sales figure lives in your order system and is usually well below the platform sum. Presenting a platform sum upward sooner or later triggers the question "why aren't the report's sales in our revenue?" — and a loss of trust. My rule: one source of truth (orders/CRM) in upward reporting; platform numbers only in their in-channel optimization context.
Which number for which decision?
All three numbers are useful — in the right job. In-channel optimization (which campaign, keyword, creative): that platform's own number. Google Ads' bidding algorithm learns from its own conversion count; arguing with it using GA4's number is pointless. Cross-channel comparison (where the budget goes): you need a single referee — GA4, or better, your own order-based reporting; what matters is that the comparison happens under one rule set. Business outcome reporting (to the boss/client): order system + total ad spend, i.e. MER / blended ROAS. Three layers, three different numbers, all legitimate — confusion begins when a layer is skipped.
Double counting you built yourself
On top of the structural gaps, setup mistakes make things worse. The ones I find most often: the same purchase firing both from a migrated Universal Analytics tag and the new GA4 event; purchase events fired on both web and server side without a transaction_id (so deduplication is impossible); the same conversion imported to Google Ads from GA4 and tracked directly by tag, with both marked primary. My audit routine is simple: place one test order and trace how many times, and with which date, it appears in every system. That one-hour test is cheaper than months of decisions on wrong data.
In short: "conversions are up" is an incomplete sentence until you say which system counted, under which attribution, to which date. Stop expecting the numbers to match, clarify what each number is for, and the same table turns from an argument-starter into a decision tool.