01 — Approach
GA4 is a shell, not a turnkey house
We have worked with Google Analytics since the Urchin days and have moved dozens of properties from Universal Analytics to GA4. The biggest lesson from that: GA4 punishes copy-paste. Rebuild the old data model one-to-one and you end up with events nobody analyzes and reports nobody understands.
That's why we don't start in the interface, we start with your questions. Which decisions should the data carry? From that comes a measurement plan, and only from the measurement plan comes the property. Anything that doesn't answer a question gets cut before it gets built.
02 — Current
What's changing in GA4 right now
GA4 is evolving noticeably in 2026. Since May, traffic from AI assistants like ChatGPT, Gemini and Claude can be analyzed as its own source. For many of our clients this has become a visible channel that nobody could attribute cleanly before. On top of that comes the new source group dimension, which collapses the tangle of variants like facebook.com, m.facebook.com and l.facebook.com into a single source. Attribution finally becomes readable.
Google is also rebuilding the interface toward planning: since February, Generated Insights on the home screen automatically summarize anomalies and changes, and cross-channel budgeting is in beta. Interesting, but handle with care: the suggestions are only as good as the conversion data underneath them.
Our take: we test new GA4 features against your setup first. An AI-generated insight on broken data is just a faster route to the wrong decision.
03 — Discipline
Setup and migration without legacy baggage
Whether it's a fresh build or the takeover of a property that grew organically: we check the foundations first. Data streams, internal filters, attribution settings, retention periods, cross-domain configuration. Sounds unspectacular, but it decides whether every number that comes later is right. For migrations we document which legacy metrics make it into the new world and which deliberately don't. That way nobody argues afterwards with numbers from two different systems.
If you are not sure where your property stands: we'll look at your setup for free and give you an honest assessment of what's solid and what needs attention. Just drop us a short note.
04 — Discipline
Event architecture with a naming convention
Events are the vocabulary of your property. We fix a naming convention before the first custom event goes live: one consistent schema, snake_case, a clear split between event names and parameters, no duplicates of the automatic events. Every event lives in the measurement plan with its trigger, parameters and owner. The technical implementation runs through Google Tag Manager, so changes stay versioned and testable.
05 — Discipline
E-commerce tracking that survives the accounting check
In shop tracking, the devil lives in the items array: missing item_ids, prices sometimes gross, sometimes net, discounts deducted twice. We build the complete chain from view_item through the checkout steps to purchase and refund, with consistent item data from a single source. We only sign off on a setup once GA4 revenue and the shop backend agree to within a few percent. Refund tracking is part of the job, because a ROAS without returns is wishful thinking.
06 — Discipline
Audiences and conversions for Google Ads
GA4 is never an end in itself for us, it's a supplier for media buying. We define key events so Google Ads optimizes on real value signals instead of cheap micro conversions, and we build audiences that actually do work in the account: cart abandoners by cart value, existing customers for exclusion, engagement segments for testing. How we put those signals to work in the account is covered on our Google Ads page.
07 — Deep dive
BigQuery export and data quality checks
The BigQuery export is free to activate and still not right for everyone. It pays off when you need raw data without sampling, want to calculate cohorts or customer journeys beyond the standard reports, or want to marry GA4 data with your CRM and shop, for example for custom dashboards. For a property whose questions the interface already answers, it's just one more place where data sits around. And if that's your case, we'll say so.
What always comes with it: ongoing data quality checks. We regularly compare GA4 against backend numbers, monitor event volumes for outliers and verify after every release that the tracking still fires. Together with server-side tracking, that keeps your data foundation stable even as browsers and ad blockers get stricter.
A number nobody regularly checks against reality is not a metric. It's a rumor.
Do you trust your GA4?
Tell us briefly what you measure and where the numbers don't add up. We'll look into your property and tell you honestly what's good and what's missing. First call free, 30 minutes.