If your attribution still rests on a browser pixel, you are already missing part of your data — you just do not know it yet. Safari cuts cookies within a day, iOS asks permission to track, and four out of five people say no. For one of our clients that looked like this in numbers: up to 40% of conversions simply never reached the report.
A lot has been written about server-side tracking and very little of it is practical. Let me try to be practical — without promising to "recover 100% of your data", because that is a lie. A great deal can be recovered. But first it is worth understanding what actually broke.
What really happened to the data
Everything used to live in the browser. A pixel set a cookie, the cookie remembered the user, and the path from click to purchase assembled itself. Then Apple shipped ITP, which cuts third-party cookies and truncates the lifetime of first-party ones. ATT followed — the "Allow this app to track you?" prompt — and the advertising signal from apps almost went to zero.
The outcome is simple. The user is still there, the purchase happened, the money arrived. But the thread linking that purchase to the advertising snapped inside the browser. The platform got no signal, could not learn from it, reported an understated result — and the algorithm started spending budget blind. The problem is not that you are "losing conversions". The problem is that you are losing the training signal for the algorithm, and that costs more.
Server-side tracking in plain terms
There is one idea: stop sending events from the user's browser and start sending them from your own server. The "purchase" event still happens on the site, but from there it no longer goes straight to Meta or Google out of a fragile browser. It goes first to a container you own — server-side GTM — and from there, clean and complete, on to the ad systems.
What that gives you in practice. The event lives inside your perimeter rather than at the mercy of browser restrictions. You control what data leaves and what stays with you. And you can send on what the browser lost — for example a confirmed purchase from your CRM that happened a week after the click.
The problem is not that there is less data. The problem is that there is less of it unevenly — and decisions get made on a distorted picture.
First-party data — the whole point of the exercise
Server-side tracking without first-party data is an expensive way to send the same truncated events, just from a server. The real value appears when you start relying on data the user gave you directly: an email at checkout, purchase history, actions in their account.
That data depends on neither Safari nor iOS, because it is yours. Hash the email, pass it through enhanced conversions, and Google and Meta can see again that this person bought, even though the cookie died long ago. Put order history into BigQuery and you get predictive segments: who will buy again, who is about to churn, who is discount-sensitive. From there you advertise not to a "lookalike audience out of thin air" but to people you actually know something about.
Server-side gets the signal back. First-party data makes that signal smarter than the ad platform.
The order to implement this in
A common mistake is starting with the hardest part. We almost always go from quick effect to deep work.
First, enhanced conversions on confirmed purchases — a few days of work and usually the most visible gain in accuracy. Then a server-side GTM container carrying key events past browser restrictions. And only then a first-party store in BigQuery with predictive segments, because that is no longer about "recovering data" but about using it better than competitors do. For Danish brand The Dybdahl Co that exact sequence lifted attribution accuracy by 40%, cut CPA by a quarter, and brought correctly attributed revenue to 95%.
Server-side tracking does not remove your consent obligations. You still need a legal basis, and analytics and advertising still require explicit consent in the EU. What changes is the technical route the data takes, not your duties under the GDPR.
It also does not fix a bad funnel. If people are not converting, cleaner data will tell you that more precisely — it will not make them convert.
How to tell whether you need this
Two checks, both quick.
- Compare purchases in your ad accounts with orders in your back office for the same month. A large shortfall on the ad-account side is the missing signal
- Look at what share of your traffic is Safari and iOS. For most European D2C brands it is substantial, and that is precisely the share where browser tracking degrades fastest
If both checks point the same way, the algorithm is spending your budget on an incomplete picture. We rebuild tracking and reporting so the numbers in the dashboard and the numbers in the accounts describe the same reality.