Google Analytics 4 Setup Guide: From Sign-Up to Reading Reports

Google Analytics 4 Setup Guide: From Sign-Up to Reading Reports

Google Analytics 4 (GA4) is the fourth generation of Google's analytics platform, released in October 2020. Its underlying model differs completely from Universal Analytics: UA was built around sessions and pageviews, while GA4 is built around events, and a single property can collect data from a website, an iOS app and an Android app at the same time. That is the real reason many people find GA4 reports unreadable at first. The interface did not get more complex; the way data is organised changed.

This guide follows the order you will actually work in: understand the concepts, complete setup and installation, configure events and key events, then read the four core report areas, with the common traps collected at the end. If you have not yet decided which tool to use, start with the comparison in Web Analytics Setup Guide and come back here to implement.

1. Understand what changed from Universal Analytics

Standard Universal Analytics stopped processing new data on 1 July 2023, and the 360 paid tier followed in October of that year, so the old reporting system has effectively retired. The most common outcome of configuring GA4 with UA habits is a set of numbers that will not reconcile.

1.1 The three levels: account, property, data stream

GA4 has fewer levels but clearer meaning. An account is the unit for permissions and billing. A property is the unit of data ownership, ideally one per business entity across platforms. A data stream is the actual endpoint that receives data, and one property can hold several website and app streams.

A practical rule: split properties by business, not by domain. Putting a brand's English and Chinese sites in one property and separating them by stream gives you a fuller picture than running two properties.

1.2 Metric definitions changed, so do not reuse old reports

• Users: the default GA4 report shows active users, not total users.
• Sessions: a new session starts after 30 minutes of inactivity or when the traffic source changes; close to UA but with different edges.
• Engagement: GA4 replaced bounce rate with engaged sessions. A session counts as engaged if it lasts more than 10 seconds, includes a conversion event, or views two or more pages.
• Pageviews: recorded as the page_view event, no longer a separate hit type.

Understand these four points and you will stop suspecting data loss every time a number looks different.

2. Setup and installation: three ways to connect

2.1 Create the account and property

Sign in to analytics.google.com, create an account with an admin identity, then create a property under it. The wizard asks for an industry category and business objectives; both only affect default report layout and suggestions, so a wrong answer does not break collection and can be edited later. Time zone and currency, however, are locked for historical data, so set them correctly the first time.

2.2 Create a data stream and get the measurement ID

Add a web data stream under the property and enter the domain; you receive a measurement ID beginning with G-. That ID identifies every tag you will install. For the same property, create separate streams for production and staging so test traffic does not pollute production data.

2.3 Option one: install gtag.js directly

Place the global snippet as early as possible in the head of every page. This is the lightest option and suits sites with a modest number of pages and stable tracking requirements.

2.4 Option two: manage everything with Google Tag Manager

When you run several sites or change tracking often, Google Tag Manager saves effort: only a container snippet goes into the page, and GA4 configuration and event tags live in the container, so tracking changes no longer require a code release. The trade-off is an extra layer to debug; validate every tag in container preview mode before publishing.

2.5 Option three: server-side with the Measurement Protocol

When you need to record actions that only the backend can confirm, such as a completed payment or an order status change, send events from the server with the Measurement Protocol. Note that server-side events need a client identifier tied to the session; otherwise you generate a large volume of unattributed traffic that actively harms attribution analysis.

3. Configure events and key events

3.1 Automatically collected events and enhanced measurement

GA4 collects events such as page_view, scroll and click by default, and enhanced measurement adds site search, form interactions, file downloads and video engagement. These switches live in the stream's enhanced measurement settings. Enable them deliberately rather than all at once, because form interaction tracking can produce noisy events on some front-end frameworks.

3.2 Custom events and parameters

Send custom events with gtag('event', 'name', {parameters}) or through a Google Tag Manager event tag. Use lowercase with underscores for naming consistency and maintain an event dictionary in advance: who sends it, on which page, under what trigger, and with which parameters. Once live, event names should follow the agreed convention, because renaming breaks historical continuity.

3.3 Mark business actions as key events

Actions such as completed registration, successful checkout or a submitted enquiry should be marked as key events in the admin area. Google renamed the former conversions report to key events in 2024; both describe the same mechanism. Only after marking them can downstream reports and advertising attribution use them.

3.4 Connect Google Ads for conversion attribution

If you also run Google Ads, link the GA4 property to the ads account so key events can be imported as conversions. This step determines whether advertising reports can show the full path from click to sale. In regions that require consent, pure client-side tracking also needs Consent Mode configured, or attribution will read noticeably low.

4. Reading reports: four areas are enough

• Realtime: verify installation and check that traffic arrives on launch day. Useful for validation, not for decisions.
• Life cycle reports: acquisition, engagement, monetisation and retention. Look at acquisition for growth, engagement for content (views, engagement time, events per user), and retention for repeat behaviour.
• User attributes and tech reports: geography, device, browser and operating system. When conversions look wrong for one region, these reports locate the device or browser layer fastest.
• Explore: free form, funnel exploration and path exploration live here. Funnel exploration is one of the most valuable GA4 features, showing exactly where users drop between a landing page and a key event. Be aware that some explore queries are sampled, so cross-check important conclusions against unsampled standard reports.

To see search-side performance such as impressions, clicks and rankings, link the Search Console property described in the Google Search Console Guide. Once linked, organic landing page performance appears directly in GA4 acquisition reports. For metric definitions and measurement conventions, see SEO and Marketing Metrics Basics.

5. Five traps worth knowing before you start

5.1 Data retention only has two settings

The free tier allows either 2 months or 14 months of retention; data older than that is aggregated or deleted. Long-term raw event storage requires BigQuery export.

5.2 Thresholding hides data in low-volume reports

To protect user privacy, GA4 applies data thresholding: when a dimension combination has too few users, those rows are withheld or left blank. If a new site or a narrow breakdown shows a total but empty detail rows, the cause is usually thresholding rather than broken tracking.

5.3 Sampling and cardinality

Large explore queries are sampled, and user-scoped custom dimensions raise cardinality, which slows processing and can trigger sampling. Be restrained with dimension design: if an event parameter can carry the value, do not promote it to a user property.

5.4 Data is not realtime

Standard reports need roughly 24 to 48 hours to settle, and realtime covers only a short recent window. Do not draw conclusions about a campaign from the same day's standard reports.

5.5 Ad blockers and Consent Mode

Script blocking means some visits are never collected, a problem shared by every client-side analytics tool. In regions such as the EU, an incorrectly configured Consent Mode removes even more attribution data. Sites with compliance obligations should plan consent management together with the tracking design.

6. Who this workflow suits

GA4 is positioned as complete, free and deeply integrated with the Google ecosystem. It fits teams that already run or plan to run Google Ads, that need to analyse web and app together, and that are willing to invest in learning the event model. If the requirement is simply to see basic traffic and sources, or if data must stay fully under your own control, compare lighter options first and then decide whether GA4 should be the primary tool.

For selection criteria, see the evaluation framework in Business Tools Overview. A side-by-side comparison of alternatives is in Google Analytics Alternatives Compared. More business tools are listed under Business Tools.

Sources: https://analytics.google.com, https://support.google.com/analytics