How GA4 models data: events, parameters, users, sessions
Universal Analytics was built around pageviews and sessions, with events bolted on as a category, action
and label. GA4 inverts that. A pageview is an event called page_view; a purchase is an event called
purchase; a scroll, a click on an outbound link, a video play and a form submission are events too.
Every event carries parameters, which are the details: the page location, the referrer, the item
purchased, the value, the form identifier. Some parameters arrive automatically with every hit, some are
set by the tag, and some are invented by the implementer for their own schema. The
events and parameters guide covers the four kinds of events and
the discipline of naming them.
Users are stitched from identifiers. On the web the default is a client identifier stored in a first-party cookie; apps use an app-instance identifier; a site that signs people in can send a User-ID; and Google signals can add signed-in Google users who have opted in to ads personalisation. The reporting identity setting decides how these are blended into one user, and the choice changes user counts.
Sessions are derived rather than sent. The tag emits a session_start event when activity begins after a
period of inactivity, every later event carries the same session identifier, and the session ends when
the inactivity timeout passes. Unlike its predecessor, GA4 does not start a new session at midnight or
when the campaign source changes mid-visit, which is one reason session counts differ between the two
systems. An engaged session is one that lasted beyond the engagement threshold, included a key event or
included more than one page view; engagement rate replaces bounce rateShare of sessions that were not engaged (GA4); historically single-page sessions. as the headline quality metric.
Three consequences follow. Reports are built by aggregating events, so a metric is only as good as the events behind it. The same event can be read at event, session or user scope, and mixing scopes produces numbers that look wrong. And because the implementer decides the schema, two properties for similar businesses can look nothing alike; the measurement plan is what keeps a property's schema tied to the questions the business asks.
Property, data streams and the admin surface
An account holds properties; a property holds data streams; a stream is one website or one app platform. A web stream has a measurement identifier that the tag uses to send events, and most sites install that tag through Google Tag Manager (GTMTag management system.) rather than by editing page templates, because the container makes later changes testable and reversible. Property settings such as time zone and currency sit above the streams; stream settings such as enhanced measurement, cross-domainConfiguring analytics so a visit that moves between a business's own domains stays one session. configuration and internal traffic rules sit on each stream.
The admin surface groups everything else under the property. Data collection and modification covers streams, data collection consent and signals, data retention, data filters and data import. Data display covers events, key events, audiences, custom definitionsA GA4 reporting field created by registering an event, user or item parameter so it appears in reports and explorations., channel groupsRules that assign each visit or conversion to a channel, such as organic search or email, from its source, medium and campaign values., comparisons, attribution settings and the reporting identity. Product links covers Google Ads, Search Console, BigQuery, Merchant Center and the other Google products a property can be joined to. Outside admin, the Reports section holds the standard reports (customisable through the library), the Explore section holds ad-hoc analysis, and the Advertising section holds attribution and conversion-path reporting. Knowing which surface owns a setting saves a large share of the time new users spend looking for things.
The core guides
- GA4 setup, step by step — property, stream, tag, enhanced measurement, retention, internal traffic, cross-domain, signals and consent, key events, product links and the checks that prove it works.
- Events and parameters — automatic, enhanced measurement, recommended and custom events; event, user and item scope; custom definitions; limits; schema design.
- Key events and conversions — choosing and marking the events that matter, counting method and value, and how they become Google Ads conversions.
- Explorations — a tutorial on free-form tables, funnels, paths, segments, overlap, user explorations, sampling and sharing.
- Audiences — conditions, sequences, membership duration, audience triggers, predictive audiences and sharing to Google Ads.
- Attribution settings — the property's model, lookback windows, channel groups and the attribution reports, as GA4 implements them (AttributionAssigning conversion credit to touchpoints. theory lives in the attribution section).
- Google Ads integration — a tutorial for linking the accounts, what flows each way and what to check afterwards.
- Reports and dimensions — the standard report collections, dimension scoping, comparisons and the library.
- E-commerce tracking — a tutorial around the documented e-commerce events and the items array.
- Common mistakes — a troubleshooting table from symptom to cause to fix for the problems that show up in most properties.
- BigQuery export basics — setting up the export, the event table schema and what it makes possible.
Further guides
Further guides apply the event model to particular analysis jobs, to people who sign in and to properties that change hands:
- Site search analysis — a tutorial for confirming the search event and query, exploring search terms, finding searches that return nothing or end the visit, and fixing them.
- Analysing A/B tests — a tutorial for sending test exposure to the property, registering the variant, checking the split, and reading the primary metric and guardrails for each variant in explorations.
- Product and merchandising analysis — reading e-commerce data for merchandising: item lists and promotions, product view-to-purchase rates, baskets and category trends.
- User-ID and signed-in users — when to send a User-ID, what value to use, how it changes user counts and reporting identity, consent rules and the join to the CRM.
- Cleaning up an inherited GA4 property — a tutorial for a property someone else built: list its settings, find what relies on each one, move Google Ads goals before anything is unmarked, fix key events and definitions, replace interface rules with tag fixes and log every change.
- Moving to a new GA4 property — a tutorial for replacing a property the business cannot own or keep: copy what only the old one holds, build the new one in the business's own account, send every event to both, rebuild what passes the clean-up tests, move links and Google Ads goals in stages, repoint dashboards and mark the break.
A sensible first month with GA4
A property that is switched on and left alone collects page views and a handful of enhanced measurement events, and that is where most properties stay. A better first month is a sequence.
- Write down the questions before touching settings. Which actions on the site mean something to the business, and which KPIA metric tied directly to an objective. each report should move. The measurement plan tutorial is the short version of this step.
- Set up the property properly with the setup tutorial: time zone, retention, internal traffic, cross-domain, consent, product links.
- Design the event schema from the questions, name events consistently, and register the parameters that reports will need as custom definitions.
- Mark a few key events, set their counting method and value, and only then link Google Ads and import them.
- Check the data daily for the first fortnight in Realtime, DebugView and the events report: missing events, duplicate events, wrong parameters and unexpected referrers appear early and are cheap to fix early.
- Build the first explorations around the questions from the first step, and turn the segments that prove useful into audiences.
A subscription box retailer's first month
A subscription box retailer launches a new GA4 property alongside its Shopify store. Week one is spent on
settings: the property is moved to the Europe/London time zone, the retention option is set to the longer
choice, the office IP range is marked internal and the filter is activated, and the checkout subdomain is
added to the cross-domain list. Week two adds the e-commerce events through Tag Manager and registers
subscription_plan as an event-scoped custom dimension. Week three marks purchase and generate_lead
as key events, with the lead event set to count once per session, and links Google Ads. By the end of
week four the property records 2,140 purchases and 980 leads, the Ads account imports both, and a funnel
exploration shows that 38% of people who reach the plan selection step leave before the address form —
the first finding the team acts on.
Common GA4 problems
Most trouble in a GA4 property shows up as one of a few symptoms. The table maps each to the guide that diagnoses it; the troubleshooting page holds the full diagnosis table.
| Symptom | Usual cause | Guide |
|---|---|---|
| Events fire but a parameter never appears in reports | Parameter not registered as a custom definition | Events and parameters |
| Key event counts far above real conversions | Counting once per event where once per session was wanted; duplicate tags | Key events |
| Google Ads and GA4 disagree on conversions | Different attribution, windows and counting time | Google Ads integration |
| Sessions restart at the payment provider | Domain missing from the cross-domain list; referral not excluded | Setup |
| Most sessions attributed to direct or a self-referral | Tag missing on some pages; redirects dropping campaign parameters | Setup, Attribution settings |
| An exploration shows a sampling warning or a thresholded row | Query exceeds the sampling quota; signals thresholds applied | Explorations |
| A high-cardinality dimension collapses into "(other)" | Standard report cardinality limits | Reports and dimensions |
| Revenue in GA4 does not match the order system | Items array or value sent wrongly; refunds untracked | E-commerce tracking, GA4 revenue not matching orders |
| A question the interface cannot answer | Needs event-level data | BigQuery export basics |
| The property shows no data at all, even in Realtime | Tag missing or unpublished, wrong measurement ID, consent or an active filter | GA4 not showing any data |
| Direct traffic far above what the business expects | Untagged links, stripped redirects, lost referrers or domain hops | Direct traffic too high |
| An audience stays empty or too small to serve in Google Ads | Conditions on events that never arrive, no backfill, link or consent limits | GA4 audience not populating |
| The shop takes orders but GA4 records no purchases, or no revenue | Buyers not returning from payment, a mistimed push, a malformed order object, missing currency, settings or consent | GA4 purchase event not tracking |
| Site search terms never appear in reports | Search event or query parameter missing, or the term parameter not registered as a custom dimension | Site search analysis |
| A test cannot be read by variant | No exposure event, or the variant not registered as a custom dimension | Analysing A/B tests |
| One signed-in customer counted as several users across devices | No User-ID sent when people sign in | User-ID and signed-in users |
| An inherited property's key events, definitions, audiences and filters that nobody can explain | Settings left by earlier owners that Google Ads, reports or dashboards may still read | Cleaning up an inherited property |
| The property sits in an account the business cannot get ownership of | An agency's or former owner's account, with no transfer on offer | Moving to a new GA4 property |
For how GA4 is read for organic search specifically — channel grouping, landing pages, engagement and organic revenue — the author's Google Analytics for SEO expertise page describes the working approach; this section teaches the mechanics.
