GA4 (Google Analytics 4) is Google’s event-based web and app analytics platform, which records every user interaction as an event with parameters and reports how people arrive at, use and convert on websites and mobile apps. In marketing analytics, it is the default free tool for measuring traffic sources, user engagement and key events such as purchases or lead submissions, and for connecting that behavior to campaigns.
GA4 replaced Universal Analytics, whose standard properties stopped processing new data on July 1, 2023. The change was not a cosmetic upgrade: GA4 uses a different data model, different metric definitions and a stronger focus on privacy, so numbers from the two generations are not directly comparable.
How the GA4 event model works
In GA4, everything is an event. A page view, a scroll, a video start and a purchase are all recorded the same way: an event name plus up to 25 parameters that describe it, such as the page location, the item value or the currency. User properties describe the person or device, and sessions are derived from events rather than being the core unit of data.
One GA4 property can collect data from several data streams, typically one web stream and separate iOS and Android app streams, which allows cross-platform reporting in a single place. Data arrives through the Google tag (gtag.js), Google Tag Manager, the Firebase SDK for apps, or the server-side Measurement Protocol.
| Event category | Examples | How it is collected |
|---|---|---|
| Automatically collected | first_visit, session_start, user_engagement | By the tag or SDK with no setup |
| Enhanced measurement | page_view, scroll, file_download, site search | Switched on per web stream |
| Recommended events | purchase, add_to_cart, generate_lead, sign_up | Implemented with Google’s standard names |
| Custom events | configurator_completed, quote_requested | Defined by the business for unique actions |
Using recommended event names and parameters where they exist matters in practice: standard reports, ecommerce views and integrations with Google Ads expect them, while custom names require extra configuration to become reportable.
GA4 compared with Universal Analytics
The main difference is the unit of measurement. Universal Analytics was built around sessions and pageviews with fixed hit types, while GA4 treats sessions as a grouping of flexible events. The comparison below explains why historical trends break at the migration point.
| Aspect | Universal Analytics | GA4 |
|---|---|---|
| Data model | Sessions with hit types | Events with parameters |
| Platforms | Separate web and app properties | Web and app streams in one property |
| Engagement measure | Bounce rate as the headline | Engagement rate, with bounce rate as its inverse |
| Goals | Goals with fixed types | Key events marked on any event |
| Default attribution | Last non-direct click | Data-driven attribution |
| Raw data export | BigQuery only in the paid tier | BigQuery export in the free tier |
Core GA4 metrics and how they are defined
GA4 metrics depend on the engaged session, a session that lasts longer than 10 seconds, includes a key event, or has at least two page or screen views. Several headline metrics are built on this definition, so understanding it prevents most reporting errors.
| Metric | Definition | Typical use |
|---|---|---|
| Active users | Users with an engaged session or engagement activity | Main audience size measure |
| Engagement rate | Engaged sessions / sessions | Quality of traffic by channel |
| Bounce rate | 1 minus engagement rate | Share of sessions without engagement |
| Average engagement time | Foreground time per active user | Depth of attention on content |
| Key events | Events marked as important business actions | Leads, signups, purchases |
| Session key event rate | Sessions with a key event / sessions | Conversion efficiency by source |
Traffic is assigned to channels by default channel groupings based on source and medium. Campaign links tagged with UTM parameters feed these dimensions directly, which is why inconsistent tagging is the most common cause of traffic landing in the wrong channel or in the unassigned bucket.
Attribution, consent and data limits in GA4
GA4 uses data-driven attribution by default, which distributes credit for a key event across touchpoints using a model trained on the property’s own conversion paths. The only rule-based alternatives left are paid and organic last click and the Google paid channels last click model; first-click, linear, time-decay and position-based models were retired in 2023.
Privacy features shape what GA4 can report. With Consent Mode, tags adjust their behavior to the visitor’s choice, and GA4 can fill gaps with behavioral modeling when traffic volume is sufficient. Some limits are structural, and analysts should know them before trusting a number.
| Limitation | Cause | Mitigation |
|---|---|---|
| Data thresholding | Small user counts hidden to protect identity | Widen date ranges or adjust identity settings |
| Short data retention | Event-level data kept 2 or 14 months for explorations | Set 14 months and export to BigQuery |
| Sampling in explorations | Large queries exceed processing quotas | Use standard reports or BigQuery |
| Unassigned traffic | Missing or malformed campaign tags | Enforce a UTM naming convention |
| Consent gaps | Visitors decline analytics storage | Implement Consent Mode correctly |
Implementing GA4 correctly
A reliable GA4 setup starts with a measurement plan, not with the tag. The plan lists the business questions, the events and parameters needed to answer them, and the events that count as key events.
- Define key events that reflect business value, such as purchases, qualified leads or trial starts.
- Deploy the tag through Google Tag Manager so changes do not require code releases.
- Use recommended events for ecommerce and leads before inventing custom names.
- Register custom dimensions for event parameters that must appear in reports.
- Set data retention to 14 months, filter internal traffic and link Google Ads and Search Console.
- Validate with DebugView before trusting any report, and document every change.
GA4 is a measurement source, not a complete reporting layer. Teams usually feed its data into a marketing dashboard alongside ad platform and CRM data, because GA4 alone cannot see offline sales or long B2B sales cycles.
A non-technical view
GA4 is a free counter and logbook for a website or app. Every time a visitor does something, GA4 writes it down. That could be opening a page, watching a video or buying a product. It also notes where the visitor came from.
Marketers then read the logbook to answer simple questions. Which campaigns bring visitors who actually buy? Which pages keep people interested, and where do they give up? GA4 does not know who individual people are. It also misses some visitors who refuse tracking. Its numbers are best for spotting trends, not for counting every single customer.
A concrete example
An online furniture retailer reviews one month of GA4 data: 120,000 sessions, of which 54,000 are engaged, giving an engagement rate of 45% and a bounce rate of 55%. The store records 1,440 purchase key events, a session key event rate of 1.2%.
Broken down by channel, paid social shows strong click volume but a low engagement rate, while a large share of sessions sits in the unassigned channel. An audit finds that the social team used inconsistent UTM values, such as several different spellings of the same medium.
After the team adopts a single UTM naming convention and marks qualified lead forms as key events, most unassigned traffic moves into the correct channels the following month. The marketing manager can now see that email delivers the highest purchase rate per session and shifts budget accordingly.