GA4 is not an upgrade of Universal Analytics — it is a different measurement philosophy. Sessions and pageviews gave way to events and parameters. Teams that map the old model onto GA4 end up with reports nobody trusts. Here is the model we install for clients, explained plainly.
Everything is an event
In GA4, every interaction is an event: page_view, scroll, click, form submission, purchase. Events carry parameters — key-value pairs like page_location, form_name, value, currency. This structure is what makes GA4 flexible, and what makes lazy setups produce garbage: if your events lack parameters, your reports lack dimensions.
The event taxonomy to implement
| Tier | Examples | Rule |
|---|---|---|
| Automatically collected | page_view, first_visit, session_start | Leave alone; verify enhanced measurement |
| Recommended events | sign_up, generate_lead, add_to_cart, purchase | Use Google's names exactly — unlocks standard reports and benchmarking |
| Custom events | whatsapp_click, calculator_use, brochure_download | Only where recommended events don't fit; document them |
| Key events (conversions) | generate_lead, purchase, phone_click | Max 5-8; everything else is diagnostic |
The discipline is in the middle rows: Google's recommended event names feed Explorations and standard reports automatically. Custom names isolate your data in spreadsheets nobody opens.
From events to key events (conversions)
Marking an event as a key event is a strategic choice — each one you add dilutes the others in reports and ad platform imports. The test: "Would we change budget based on this number weekly?" If yes, it is a key event. A lead form submission qualifies; a 50% scroll almost never does.
Parameters that make reports useful
- lead_source / form_name — separates campaign landing forms from footer forms.
- service_interest — segments leads by intent.
- value and currency — enables revenue reporting even for non-ecommerce.
- traffic-quality markers — feed your CRM stage back as a parameter for closed-loop reporting.
The setup path
- Create the property with a clean data stream per environment.
- Implement via GTM (see our GTM guide) with a data layer.
- Configure enhanced measurement, then verify against server logs.
- Mark key events, import them into ad platforms, and annotate launch date.
- Build one Exploration that answers your weekly business question — before building ten that answer nothing.
Mistakes that silently break GA4 data
- Duplicate page_view from native collection plus a GTM tag.
- Consent mode blocking everything in one region and nothing in another — making regional comparisons meaningless.
- No internal traffic filter — your team becomes your best-performing "campaign."
- Dimension Cardinality explosions from stuffing unique IDs into parameters.
- Never auditing: events drift as the site evolves.
Do I still need GTM with GA4?
Yes. GA4's native tags cover basics; GTM gives you version control, testing, and a single place where all platforms consume the same data layer.
How many key events is too many?
Beyond eight, attention fragments and ad platforms receive noisy signals. Five well-chosen key events outperform fifteen casual ones.
Measurement architecture is a specialty of ours — GA4 implementation with GTM, data layer design, and reporting that answers business questions. Talk to us.