Pixel, Google Tag Manager and Consent Mode: tracking conversions correctly

You pay for ads on Facebook and Google. At month end you look at the reports and something doesn't add up: Meta claims 60 orders, Google says 40, GA4 shows a different number again, and the shop recorded 70 sales in total. Who is lying? Usually nobody. Each platform simply counts something different, and you are making budget decisions on numbers you have no right to trust.

The problem is almost never the ads. It is the layer beneath them — conversion measurement. The pixel installed wrong, firing twice or not at all. The code pasted in "to make it work" three years ago that has been quietly losing signal ever since. And a new context: third-party cookies are disappearing, iOS and Safari restrict tracking, and Consent Mode became mandatory for anyone advertising to EU users. The result: old setups lose between a fifth and two fifths of conversions without telling you.

This article is about setting measurement up correctly, once, so the numbers mean something.

Why the numbers no longer match

For ten years, measurement ran on a simple principle: the visitor's browser runs a piece of code (the pixel), which drops a cookie and reports back to the platform. That model is collapsing from three directions at once.

• Third-party cookies are disappearing. Safari has blocked them for years, and the other browsers are heading the same way. Without them, the platform cannot connect today's visit to the ad someone saw three days ago.
• iOS asks the user. Since App Tracking Transparency, a large share of people decline tracking, and the platform is left with no data on them.
• Ad blockers and extensions stop the pixel before it ever fires.

Every lost signal is not just an unreported order. Meta's and Google's algorithms learn from conversions — who buys, so they can find more people like them. With half the signal, optimisation learns half blind, and you pay more for the same result. That is why measurement is not a matter of "pretty reporting" — it is money straight off your ad invoice.

The parts of the system, briefly

To keep the jargon in check, here is what each piece does:

• The pixel / tag (Meta Pixel, Google tag) — the code on the page that reports events: product view, add to cart, purchase.
• Google Tag Manager (GTM) — a "control room" where you keep every tag in one place and manage it without touching the site's code each time.
• The Conversions API (Meta) / server-side tracking — sends events from your server rather than the browser, to recover the signal the browser loses.
• Consent Mode — the mechanism that makes Google's tags behave differently depending on what the visitor accepted in the cookie banner.

Google Tag Manager: one place for everything

The most common mistake we find in audits is the chaos of codes pasted directly into the site: a Meta pixel here, Google Ads code there, GA4 somewhere else, each added by a different person in a different year. Nobody knows any more what fires, when, or how many times. Duplicates are the rule, not the exception — and an order counted twice breaks every return-on-ad-spend calculation.

GTM fixes this by putting a single piece of code in the site, while every other tag is managed from one interface. The concrete benefits: you add new tracking without a developer, you see in a preview mode exactly what fires on each page, and you have one place to clean up when you want order. For a store, this is where you configure the events that matter — add_to_cart, begin_checkout, purchase — with the order value attached, so the platforms optimise for revenue rather than click count.

Consent Mode: the part you cannot skip

Since March 2024, Google requires Consent Mode v2 of any business using Google audience features (remarketing, custom audiences) with users in the European Economic Area. In short: if the tags detect EU users and do not receive the consent signal in the required format, Google stops collecting data for those users.

"But I'm in Moldova, not the EU" — this is the trap. Most of our clients advertise into Romania, which is in the EU. The moment your targeting includes European users, the requirements apply. Without Consent Mode set up correctly, you risk two things at once: your remarketing audiences empty out (Google is no longer allowed to add people to them) and you stay exposed on the legal side.

Consent Mode works together with the cookie banner: the banner asks, and its signal tells the tags what they are allowed to do. When the visitor declines, Google sends anonymous pings without cookies and statistically estimates the lost conversions — that is how you recover part of the data, legally. The official documentation for the mechanism is at developers.google.com. The banner and legal-basis side ties directly to the obligations we covered in our guide to data protection on your website.

The Conversions API: recovering the lost signal

This is the difference between a 2020 setup and one that holds up in 2026. The classic pixel sends events from the visitor's browser — exactly where blockers, iOS and Safari cut the signal. The Conversions API (at Meta) and server-side tracking send the same events from your server, where nobody blocks them.

Ideally you run both in parallel: the browser pixel and the server, each event carrying a unique event_id. The platform sees it is the same order, counts it once (deduplication), and fills the browser's gaps with what it received from the server. The typical result we see after a clean implementation: 15–30% more conversions counted correctly, which means better-fed algorithms and a lower cost per acquisition. You are not inventing sales — you are seeing the ones that were happening anyway but weren't being counted.

How to check what you have now

• Does the pixel fire exactly once? Meta's and Google's preview tools show every fire in real time.
• Does the purchase event carry a value and currency? Without them, value-based optimisation does not work.
• Is Consent Mode configured? Check that the banner actually sends the signal to the tags, not just displays a button.
• Are you running the Conversions API? If not, that is the first source of recoverable signal.
• Are GA4 and the ad platforms in the same range? A 10–20% difference is normal (different attribution models); a 2–3x gap means something is broken.

These numbers are the foundation of every marketing decision. There is no point optimising campaigns in Meta Ads or Google Ads if the data feeding them is wrong — and the reports in Google Analytics 4 are only correct when the measurement layer beneath them is correct.

Conclusion

Measurement is the boring part of digital marketing, and that is exactly why everyone puts it off. But every budget decision you make — more here, less there — rests on numbers. If the numbers are wrong, you optimise in the wrong direction and pay for it month after month.

The good news: it is a setup you do once. A clean GTM, the Conversions API switched on, Consent Mode configured legally — and from there the numbers mean something. Start with the check above. If you find pixels firing twice or events with no value, you already have the first explanation for why the ads "don't work" the way the reports claim.