Your tracking is broken and you're making decisions on half the data

On a ready-to-wear brand we took over recently, we pulled the numbers right at the start. 25,000 clicks on the campaigns. And only 12,000 pageviews coming through. Half of it disappearing into thin air.

When I see that, the first reaction is to assume the tracking is broken. And often, it is. The problem is that as long as you keep your campaigns running on that, you're making every decision with half the info. You scale a budget when you shouldn't. Or worse, you kill a campaign that was working perfectly fine but whose conversions just weren't coming through. The worst part is it shows up nowhere in your reports, and all the while it's skewing every decision you make.

Why 50% loss isn't necessarily a bug

Let's add some nuance here. When you see 25,000 clicks and 12,000 pageviews, your first instinct is to scream broken tracking. But sometimes it's just the cookie banner.

What actually happens is people click your ad, land on the site, and a big chunk of them refuse cookies. So the pageview never fires on the tracking side. On some accounts you might have only 30% of people accepting the banner. That's not a bug, it's just how consent works today.

So the real question is never really "am I losing data": yes, always. What you need to look at is whether you're losing more than normal, and how much of it you can recover.

On this brand, 50% loss was above what we usually see. So there was a genuine technical issue behind it, not just consent. And when Meta only has half the info to learn from, the algo is working blind. It optimizes, but badly.

The emergency checklist when taking over an account

The second we start a collaboration, tracking is the top priority. Before we even touch the campaigns. Because otherwise we'll make mistakes and miss opportunities, and everything we say about paid will be built on sand.

Here's what we check first:

• Meta's Conversion API. This is what lets you send events server-side and recover a chunk of the data you lose with the browser and refused cookies.
• UTMs. Are they clean, consistent, can you actually trace where each visit comes from.
• GA4. On this particular brand, it wasn't even installed. We set it up, but until it's been running for a while, reliability isn't 100% guaranteed.
• The end-to-end test. You place an order on Shopify, and you check it comes through properly in Meta AND in Google.

That last point is the dumbest and the most important one. You make a real purchase, and you follow the path: ad, page, cart, order. If the order doesn't land in Meta, you've found your leak. No need for a fancy tool to do this.

Something that really helps with data loss is server-side tracking. Instead of routing everything through the customer's browser (with refused cookies, blockers, etc.), you push the data server-side. You recover part of what was getting lost. It's fairly quick to set up and it changes the quality of what Meta and Google receive.

The source of truth trap

Here's something that drives people crazy, and rightly so. You look at your add-to-cart rate on Shopify, it tells you one thing. You check it on GA4, it tells you something else. You go to Meta, a third number entirely.

And now you're wondering which one to trust. Honestly none of them is "the right one", each platform counts in its own way, with its own attribution window and its own method. And if you spend your time comparing three numbers that all move, you make no decisions, you just have pointless debates in meetings.

The rule I apply: you pick ONE source of truth and you stick to it. We use Shopify for revenue and actual orders. That's the money that genuinely comes in. Then, to judge campaign performance, I look directly at the platforms, the Meta view on Meta, the Google view on Google. That way I don't get an attribution bias mixing everything up.

The goal was never to have the perfect number. What you want is to ALWAYS keep the same reference, so you compare apples with apples from one month to the next. If you want to dig into the underlying logic, we also cover it in this article on why an audit should start with your P&L and not your campaigns.

The product feed, fashion's forgotten tracking

Everyone talks about pixels and events, but there's another place where things break silently: the product feed. The one that powers Google Shopping and the dynamic carousels on Meta.

In fashion, it's critical. You've got dozens of references, size and color variants, collections that turn over fast. If your feed is dirty (badly written titles, empty categories, images that don't follow through, out-of-stock products still active), you burn budget on products nobody can buy, and Google shows you less because it rates your feed as low quality.

A clean feed isn't sexy to show off in a meeting. But in fashion eCom, it's a big part of what makes paid shopping run.

The case of the English site with French ads

A real in-house example that hurts. A site in English only, with Google ads running in French. The result: the campaigns got heavily penalized.

The reason is simple: Google looks at the consistency between the ad and the landing page. Your ad is in French, the visitor clicks, they land on a page entirely in English. Google rates relevance as poor, your Quality Score drops, your costs go up and your delivery falls. It's not a manual penalty, it's mechanical.

So before you push budget, you check that the ad's language matches the landing page's language. Sounds obvious when you say it like that. In reality, it slips under the radar all the time. It's the kind of detail you catch when you stop trying to control everything on Meta and go look at what's actually happening behind the click.

Where to start

For us, tracking and feeds get set up right at the start, in parallel with the audit. The audit takes about 2 weeks, and during that time we're already tackling the data work. No way we wait around to get clean data.

If you take away one thing: the more reliable your data, the better the algo works. It's as simple as that. Meta and Google learn from what you send them. Send them incomplete data, and optimization follows suit: they work with what they've got.

Before tweaking a targeting or scaling a budget, go check that a real order comes through everywhere. If that test fails, everything else is just noise.