1 · GA4 no longer measures, it models
When a visitor refuses cookies, GA4 does not really count them. It estimates them. Since Consent Mode (Google's system for cookie consent), an artificial intelligence reconstructs the behavior of visitors who said no, from those who said yes. Google calls it behavioral modeling.
The principle is documented in black and white by Google: the model trains on consenting users to guess the others (Google Analytics Help). Yet consenting users convert 2 to 5 times more than non-consenting ones (Google Ads Help). So GA4 extrapolates a population that behaves differently from the one it observes.
These modeled conversions land in the same column as the real ones, with no label. You read a figure. You do not know which share is counted and which is estimated. For a marketing lead, that is the blind spot: you steer a budget on data whose truth rate you do not know.
2 · What vanishes before it even arrives
Modeling handles cookie refusals. That leaves everything that never fires the tag. Nearly 30% of users, up to 43% depending on the dataset, use an ad blocker (Backlinko). Many of those blockers also cut measurement scripts, GA4 included. The signal leaves before it arrives. You do not even model it, you do not see it.
Add browsers. Safari and Firefox have limited the lifespan of measurement cookies for years. A visitor who comes back ten days apart can be recounted as two people. Your acquisition source, meanwhile, gets lost along the way.
This is not a small margin gap. It is a structural part of your traffic that never enters the tool. On the accounts we audit, the gap between real baskets at the till and what GA4 tells you runs into tens of percent, not points.
3 · The cookiepocalypse myth, let us set the record straight
For five years you have been told third-party cookies will die and you must change everything before the deadline. That is false, and it matters to say it. In July 2024, Google gave up removing third-party cookies from Chrome. In April 2025, it confirmed: no removal, not even a choice screen forced on the user (Forrester).
So no, cookies are not dead. The real problem was never their sudden disappearance. It is their slow degradation and fragmentation: a piece blocked by ad blockers, a piece shortened by Safari, a piece estimated by Consent Mode, a piece still intact on Chrome. You no longer measure a reality, you measure a mosaic of pieces of different reliability.
That is trickier than a clean death. A clean death you see, and you adapt. A slow degradation you suffer without noticing, while still reading your numbers as if they were worth those of five years ago.
4 · The trap we see most: your year minus one no longer means much
Here is the point nobody raises, and it is the costliest in decisions. The share of data that escapes you is not stable over time. Ad blocking grows. Browsers tighten the screw every year. Consent Mode models a share that moves with your consent rate.
So comparing your quarter to last year's amounts to comparing two different measurement systems, not two quarters. Your minus 15% year over year may be a real drop in business. Or you just capture less signal than before. You cannot tell with GA4 alone.
For a marketing lead, this is the most expensive trap. You make budget decisions on a year-over-year comparison. If the ruler that measures deforms between the two points, the comparison lies, even when each figure taken alone looks clean. The only guard: a stable measurement base you control, not a tool whose method shifts under your feet.
5 · Why you keep GA4 anyway
All of this might make you want to throw GA4 out. Bad idea. Its value was never the precision of its measurement, the one we just took apart. Its value is the Google ecosystem it is wired into. On that ground, nothing replaces it.
First pillar: Google Ads. GA4 feeds your conversions and audiences straight back into Google Ads. Smart Bidding, the automatic bidding that drives your campaigns, learns on those signals. No third-party tool feeds Google's bidding machine as cleanly as its own analytics. Cutting GA4 starves your paid acquisition on Google.
Second pillar: the Search Console (search.google.com/search-console), Google's free tool that shows your organic search queries. GA4 ties into it. It is the only clean bridge between what people type into Google, the positions you hold, and what they do next on your site. Without GA4 in the loop, you lose that link between organic search and behavior.
So the right framing is not "GA4 or something else." It is "GA4 for what it does alone." You keep it as the hub that talks to Google's machines, you stop treating it as an accountant, and you add the rest on top. The ecosystem is what saves GA4, not its raw data.
6 · Take the signal back at the source: server-side
Server-side tracking means sending events from your server, not from the visitor's browser. On the browser side, ad blockers cut, Safari shortens, the network drops. On the server side, you control the send. You recover a share of the signal that was evaporating.
In practice, you go through a server-side container, hosted by you or a provider. Tools like TAGGRS set up that container without you managing the infrastructure, for a few dozen euros a month depending on volume, plus one to three days of tuning. It is not reserved for large accounts. A serious e-commerce SME can reach it.
One thing the hype on the topic forgets to say: server-side does not bypass consent. You stay bound by GDPR. What you gain is technical reliability on the signal of visitors who did, themselves, agree to be measured. It is documented in our server-side tracking, and it is the first brick of measurement that holds.
7 · Your platforms already know more than GA4
While you fight to reconstruct your conversions in GA4, your e-commerce platform already knows them. Shopify, in its back office, knows exactly how many orders came in, for what amount, for which customer, with which product. That data is first-hand, it depends on neither cookies nor ad blockers. GA4 only recounts it from outside, less well.
The reflex to build: cross your platform data with your marketing measurement, instead of running everything through a single browser-side tool. Your revenue reads in your back office. Your attribution, GA4 can help you light it up, but it is no longer the judge.
For on-site behavior, a first-party analytics like PostHog keeps the data with you rather than at Google. Set up in respect of consent, it is what we wired ourselves for the HeySquad site. You see the journey, the friction, the drop-offs, on your own data. Here too it is a complement, not a replacement: GA4 stays useful for volume and trend.
8 · The real 2026 move: own your data
The differentiator, in 2026, is no longer the tool you pick. It is your ability to own your raw data, and to know how to use it. Good news: the first step is free and few people take it.
GA4 lets you export every raw event to BigQuery, Google's data warehouse. The export is free for a standard property, with no sampling, without the 14-month limit of the interface, and the data is yours (Google Analytics Help). You only pay for storage and queries beyond a generous free tier. That is your first datalake, the lake where you centralize your raw data to cross it your way.
The datalake is no longer reserved for large accounts. What is most often missing is not budget, it is skill. Knowing how to read raw data, write a query, cross your back office and your measurement. Upskilling on that, or surrounding yourself, pays more today than yet another bought dashboard. Strategic data has become an asset you build, not a report you consume.
Where to start, without re-tooling everything: keep GA4 for the Google ecosystem, its links with Google Ads and Search Console, add server-side to make conversion reliable, read your platform for the real figure, and turn on the BigQuery export now to start owning your raw data. It is gradual, not a big bang.
