Google Analytics 4A working guide for marketers

How GA4 thinks: events, not pageviews#

GA4's core idea is that everything is an event. A page view is an event, a scroll is an event, a form submission is an event, a purchase is an event. Each event can carry parameters: extra detail such as the page location, the value of a purchase or the name of a button.

This is a better model for modern journeys that span web and app, but it moves effort to the front. In older analytics tools much was inferred for you. In GA4 you decide which events matter, how they are named and which parameters travel with them. A property set up carelessly will produce reports that look complete and mean little.

Sessions and users still exist, but they are derived from events. That is why session counts in GA4 rarely match an older tool's: the counting logic is different, not necessarily wrong. Compare trends within GA4 rather than across tools.

GA4 also replaced bounce rate as the headline quality measure with engaged sessions: sessions that last beyond a minimum time, include a key event or view more than one page. Bounce rate still exists as the inverse, but engagement is the more useful lens because it rewards pages that answer a question quickly.

The four kinds of events#

GA4 groups events into four families, and knowing which family an event belongs to tells you how much work it needs. Google's documentation lists the current events in each family; check it before naming anything yourself.

Our rule is simple: use a recommended event wherever one fits, because GA4's reports and integrations understand them. Create a custom event only when no recommended name describes the action. Our detailed walkthrough of GA4 events and conversions covers naming and parameters.

A sound setup, step by step#

Most GA4 problems trace back to the first hour of setup. The sequence below is the order we would follow on a new property, and the order we would audit an inherited one.

Two settings deserve special mention. Data retention controls how long user-level data is kept for explorations; the default is short, and many properties should extend it to the longest option their plan allows. Internal traffic filters stop your own team from inflating numbers. Both are easy to forget and painful to fix after months of data.

Deploy through Google Tag Manager rather than hard-coding where you can. It keeps tracking changes out of release cycles and makes consent mode easier to manage.

Key events, conversions and why the words changed#

GA4 lets you mark certain events as the outcomes that matter. Google has renamed this concept over time: what GA4 once called conversions it now calls key events, while 'conversions' is used for outcomes shared with Google Ads for bidding. Terminology may change again, so check the current documentation, but the principle is stable.

Mark only events that represent genuine progress for the business: a purchase, a qualified enquiry, a booked demo, an account creation. Marking every scroll and click as a key event makes the report meaningless and, if imported into ad platforms, teaches bidding algorithms to chase cheap actions.

Which reports to use, and which to treat carefully#

GA4 has two working areas. Reports are pre-built, aggregated and good for routine monitoring. Explore offers free-form analysis: funnels, path exploration, segment overlap and cohorts. Explorations are where the diagnostic work happens, and where data retention and thresholds matter most.

Treat two things with care. Thresholding can hide rows when user counts are small, particularly where certain identity features are enabled. Modelled data may fill gaps left by users who declined consent, depending on your setup. Neither is a flaw, but both mean GA4 figures are estimates. For ecommerce, revenue in your order system remains the source of truth.

Connecting GA4 to everything else#

GA4 becomes far more useful when linked. The Google Ads link lets you share audiences and import key events. The Search Console link adds organic query data (see our Search Console guide). The BigQuery export gives you raw, event-level data you can join to CRM and cost data, which is the foundation for serious marketing dashboards.

For teams with any analytical ambition, we treat the BigQuery export as close to essential. The interface is a convenience; the raw data is the asset. Check Google's current documentation for export limits and costs on your plan.

The mistakes we see most#

Two quieter mistakes cost more over time. The first is never reconciling: nobody checks GA4 purchases or leads against the order system or CRM, so drift goes unnoticed for months. The second is letting every team create its own events and dimensions until the property is a thicket of duplicates. A short event dictionary and a monthly reconciliation prevent both.

The deeper mistake is treating GA4 as the measurement strategy. It is one instrument. It tells you what happened on your properties, reasonably well. It does not tell you what your marketing caused; for that you need incrementality testing and, at scale, modelling.

Published by Fabulous.Media, a network of specialist marketing agencies. Updated 9 October 2026. Platform features change often; check current official documentation before acting on platform-specific detail.