Why Meta, Google and GA4 All Report Different Numbers

Ask Meta how many sales it made last month, then ask Google Ads, then open GA4, then check your CRM. Congratulations: you now have four different answers, and every one of them is delivered with total confidence.

Here’s the part nobody tells you. That’s normal. It isn’t a bug, or a sign someone set it up wrong. Your platforms disagree because each one counts a sale by its own rules, and they were never designed to match. The job isn’t to force one clean number out of them. It’s to understand why they differ and which view to trust for which decision.

This is one of the four numbers I reckon every business owner should be able to answer, and it’s the one that causes the most quiet panic, so it gets its own guide. If you want the wider picture first, start with the four marketing questions every owner should be able to answer. This piece is about marketing attribution: why the dashboards fight, and how to read four of them without losing the plot.

Why the platforms can’t agree, and never will

Marketing attribution is just the question of who gets credit for a sale. When someone buys, which ad, click or channel gets to claim it? Every platform answers that question differently, and they all answer it in their own favour.

Meta counts a sale if that person saw or clicked a Meta ad inside its attribution window, which includes people who never clicked at all, only scrolled past. Google Ads counts the sales it can tie back to a click it charged you for. GA4 uses its own model and only sees what its tag manages to track, so it usually credits ads with less than the ad platforms claim for themselves. Same sale, three different sets of rules.

Then there’s the obvious conflict of interest. Every ad platform is also trying to sell you more ads, so each one is quietly motivated to take credit for as much revenue as it can. That’s why a single sale can show up in Meta’s dashboard and Google’s dashboard and GA4 all at once, as if you made three sales instead of one. Nobody is lying. They’re each answering “did I help?” and all saying yes, which is also why platform-reported ROAS almost always looks rosier than your bank account.

What a real gap looks like: 835 orders, 675 tracked

Here’s one from last month, so you know this isn’t theory. We run the ads for an ecommerce brand that sells worldwide, and their Shopify showed 835 orders while GA4 counted 675 purchases. On paper that’s 160 sales gone missing, about a fifth of the business, somewhere between checkout and dashboard.

Before anyone panicked, we did the boring, important thing and checked whether the two were even measuring the same event. Shopify records every paid order, full stop. GA4 only logs a purchase when its tag actually fires, and a big chunk of shoppers, especially in consent-heavy markets like the EU, reject analytics cookies, block trackers or browse private. The sale still happens, GA4 just never hears about it.

The tell that it wasn’t a broken tag was almost funny. Shopify said orders were down about 10% on the month, and GA4 said purchases were down 9.9%. That near-perfect match meant GA4 hadn’t suddenly broken, it’s been quietly catching about 81% of purchases all along.

So what did the client actually have? 835 orders and $329,669 in Shopify, not the 675 and $257,000 GA4 was showing. I’d never hand them “675 sales” as their result. Shopify’s the source of truth for what they banked, and GA4 still earns its keep, just for different jobs: trends over time, landing-page performance, which channels are pulling, how conversion rate moves. You read it for direction, not for the final tally.

A dashboard shows data, it doesn’t make it true

Plenty of businesses tell me they’ve got attribution sorted because they’ve got a Looker Studio dashboard. I love the confidence, but a dashboard only displays data. It doesn’t check whether that data is any good. You can build the most beautiful report known to humankind and fill it entirely with rubbish.

Real attribution starts much further down the chain than the chart. It’s about whether each stage of the journey actually talks to the next one: ad click, website visit, conversion, lead, CRM, qualified opportunity, customer, revenue. If Google Ads says it drove 60 leads but your CRM only holds 42 genuine enquiries, that gap is the interesting part, not something to paper over with a nicer graph. Getting those stages to line up is the real work, and it’s the same discipline behind honest agency reporting and measuring paid ads properly.

The one customer journey that breaks every model

Here’s a completely ordinary path to a sale. Someone sees your Instagram ad and doesn’t click. Three days later they Google your business, read a blog, and leave. A week after that they see another Meta ad, visit your profile, then Google you again, click your search ad, and finally buy.

So who won that customer? Meta? Google? Organic search? Instagram? The honest answer is probably all of them, in sequence. This is where attribution models come in, and why they hand you different numbers. Last-click gives every dollar of credit to that final Google search. First-click hands it all to the original Instagram ad. Linear splits it evenly, and a data-driven model carves it up by influence. None of them is the truth. They’re all just opinions about how to divide one messy story, which is exactly why running your business off a single attribution number is a trap.

See how each model splits one sale

Same journey, four attribution models. Tap one and watch the credit jump between channels.

The four views worth running

So don’t run the business off one dashboard. Keep four open, and use each for the job it’s genuinely good at.

None of these four is the winner. Each answers a question the others can’t: platform data tunes the campaigns, GA4 shows the on-site story, the CRM tells you which leads turned into money, and the blended view keeps everyone honest about whether the whole thing is working.

From the studio
The panic call I get most sounds like this: “Meta says it made us forty grand but Shopify only shows fifty-five total, something’s broken.” Nothing’s broken. Meta is telling you how much revenue it thinks it touched, and Shopify is telling you what the whole business actually banked. Two honest answers to two different questions. My job isn’t to make those numbers identical, it’s to help you understand the gap and stop making decisions as if only one of them is real.

How to read four dashboards that disagree

Once you accept the numbers won’t match, reading them gets a lot calmer. A few habits do most of the work.

Anchor on the blended view. Total marketing spend against total revenue, for the whole business, over the same period. It’s the one number no platform can flatter, and it’s the closest thing you have to the truth. MER, your marketing efficiency ratio, is exactly this, and it’s the sanity check I’d never run a budget without.

Let each platform do its own job. Use Meta’s and Google’s numbers to optimise inside those platforms, not to referee against each other. They’re good at telling you which ad is pulling its weight, and hopeless at telling you who really deserves the sale.

Make your tracking trustworthy, not flawless. Clean conversion tracking, proper UTMs on your links, and a CRM that records where each lead came from will answer the questions that actually change your decisions. Perfect isn’t on the menu, and chasing it burns money you could spend on the ads themselves.

How much attribution do you actually need?

Here’s the reassuring bit. Most businesses need far less measurement than the internet implies. The right amount scales with the size of the decisions you’re making, not with how much data you could theoretically collect.

If you’re a local trades business spending $3,000 a month on Google Ads, you don’t need a data warehouse and a fourteen-touch model. You need accurate conversion tracking, call tracking, a clean CRM, proper UTMs, simple lead-to-sale reporting, and clear visibility over what you spent and what came back. That’s enough to answer the questions that matter, and it’s usually where the leaks hide too, which is the same reason ads get clicks but no sales.

Bigger budgets and bigger bets earn more sophisticated measurement: incrementality testing, media mix modelling, the proper machinery. But you graduate to that when the decisions get harder, not because a dashboard looked empty. The goal was never to collect every data point humanly possible. It’s to collect enough accurate data to make better calls, and to stop treating four disagreeing numbers as a problem when they’re just four honest angles on the same sale.

Frequently asked questions

Why do Meta and Google report different sales numbers?

Because each one counts conversions by its own attribution rules and window, and each is motivated to claim the sales it influenced. The same purchase often gets counted in Meta, Google and GA4 at once. They’re not wrong, they’re answering slightly different questions about the same sale.

Which attribution number should I trust?

None of them on its own. Anchor on your blended view, total spend against total revenue, then use the platform, GA4 and CRM numbers to understand the detail underneath. One dashboard should never run your whole business.

Why does GA4 show fewer conversions than Google Ads or Meta?

GA4 usually uses a different model and mostly ignores view-through, where someone sees an ad without clicking. The ad platforms happily count those, so they credit themselves with more conversions than GA4 will hand them.

What is marketing attribution, in plain terms?

It’s how you assign credit for a sale across the touchpoints that led to it, like ads, search, social and email. Different attribution models split that credit differently, which is exactly why the numbers move around depending on where you look.

Do I need perfect attribution to run good marketing?

No. You need tracking that’s trustworthy enough to make decisions from. For most small businesses that means clean conversion tracking, proper UTMs and a tidy CRM, not a data warehouse and a full-time analyst.

Can I make all the platforms match?

No, and trying to is a waste of time and money. They’re built to count differently. The goal is to understand why they differ and read them together, not to force one identical number out of them.