A Complete Guide to Google Analytics 4

Google Analytics 4 is now the only version of Google Analytics. Universal Analytics stopped processing new data on 1 July 2023, and Google removed all access to UA properties on 1 July 2024, which means every site still using Google Analytics in 2026 is running GA4. This guide covers how GA4 actually works in 2026, the changes that came after the original launch, and the setup choices that produce useful reporting versus the defaults that quietly under-report.

Key Takeaways: GA4 is the only Google Analytics in production. Universal Analytics was permanently removed on 1 July 2024. GA4 is event-based, cross-device, and built around the Chrome cookie-deprecation reality. Consent Mode v2 is mandatory for EU traffic. Free BigQuery export is one of GA4’s biggest under-used features. Set up properly, GA4 is genuinely better than UA was.

What is Google Analytics 4 and how is it different from Universal Analytics?

Google’s GA4 documentation describes GA4 as an event-based analytics platform that measures interactions across websites, apps, and devices in a single property. Universal Analytics was session-based; GA4 is event-based, which is the single biggest structural change between the two.

Practical differences that matter:

• Event model. Every interaction in GA4 is an event with parameters. UA used sessions, pageviews, and event categories.
• Cross-device by default. GA4 stitches web and app data in one property using Google signals and user IDs.
• Privacy by design. GA4 ships with cookieless behavioural modelling, consent integration, and built-in IP anonymisation.
• Predictive metrics. Purchase probability, churn probability, predicted revenue, built into the platform.
• Free BigQuery export. GA4 streams raw event data to BigQuery for free, where UA charged for it.
• Different reporting interface. Reports are now configurable, with explorations and a separate “Reports” section.

The transition was painful for teams who had years of UA history they could not import. Teams who set up GA4 alongside UA in 2022 and 2023 now have years of comparable data. Teams who delayed are starting their measurement history from scratch in 2024 or 2025.

How do you set up Google Analytics 4 correctly in 2026?

Google’s GA4 setup assistant walks new properties through the basics, but the default setup misses several configuration steps that materially affect data quality. A correct GA4 setup covers data streams, enhanced measurement, custom events, conversions, and BigQuery export.

A working setup checklist:

1. Create the property and data stream. Web for a website, App for an iOS or Android app, or both for a cross-platform business.
2. Enable Enhanced Measurement. Page views, scrolls, outbound clicks, site search, video engagement, file downloads, form interactions. Auto-captured.
3. Set up Google Tag (gtag.js) or Google Tag Manager. GTM is more flexible; gtag is simpler.
4. Mark conversion events. Form submits, purchases, demo bookings, key page views. Up to 30 conversion events per property.
5. Link Google Ads and Search Console. Acquisition data and ad-cost-to-revenue flow through cleanly.
6. Enable BigQuery export. Free, and unlocks SQL-level analysis the GA4 UI cannot do.
7. Configure data retention. Default is two months for event-level data; bump to 14 months.
8. Configure cross-domain tracking if the user journey spans multiple domains.

The setup-assistant migration from UA to GA4 ran from 2022 to 2024 and is no longer relevant in 2026. New properties start clean.

What is Consent Mode v2 and why does it matter?

Google’s Consent Mode v2 documentation describes a privacy-by-design framework that adjusts how GA4 tags behave based on visitor consent. In the European Economic Area, Switzerland, and the United Kingdom, sites running Google services for personalised advertising must implement Consent Mode v2 to remain eligible to use audience and conversion features.

Consent Mode v2 in practice:

• Two new consent signals. ad_user_data and ad_personalization, in addition to the existing analytics_storage and ad_storage.
• Basic mode. Tags do not load until consent is granted; no data sent if denied.
• Advanced mode. Tags load and send anonymised “cookieless pings” if consent is denied. Behavioural modelling fills the gaps.
• Required for EEA audiences in Google Ads. Non-compliant properties lose access to personalised audiences.
• Compatible with most CMPs. Cookiebot, Iubenda, OneTrust, Usercentrics, and others support the v2 signals.

Many sites still run a CMP (consent management platform) that only handles the GDPR cookie banner but does not pass the Consent Mode v2 signals back to Google Tag. If your CMP is more than 18 months old, audit whether it actually sends the four required consent signals. The cookie banner can be perfect and the GA4 reporting can still be broken because the CMP is not signalling consent back to Google.

How does GA4 reporting actually work?

Google’s GA4 reports documentation describes a three-layer reporting architecture: the standard Reports section for routine monitoring, the Explore section for custom analysis, and the Advertising section for ad-spend integration. Each layer answers different questions.

The reports most teams use day-to-day:

• Acquisition (Reports section). Where visitors come from: organic, paid, referral, social, direct, email.
• Engagement (Reports section). What visitors do: pageviews, events, conversions, time on page.
• Monetization (Reports section). Revenue, transactions, AOV, ecommerce funnel.
• Retention (Reports section). New vs returning, cohort behaviour.
• Exploration (Explore section). Funnel analysis, path analysis, segment overlap, user lifetime value.
• Advertising Workspace (Advertising section). Attribution modelling, conversion paths, Google Ads ROI.

GA4’s reporting feels less prescriptive than UA’s. The exploration section is more flexible but has a steeper learning curve. Teams who never moved past the default reports rarely get GA4’s full value.

What are GA4 predictive metrics and how do you use them?

Google’s predictive audiences documentation describes machine-learning models built into GA4 that predict future user behaviour: purchase probability, churn probability, and predicted revenue. These metrics power audience definitions in Google Ads and let you target users likely to convert, not just users who already converted.

Predictive metrics available in GA4:

The predictions require minimum data thresholds (typically a few hundred users with the target behaviour per week) before they activate. Once active, they work well for both audience building in Google Ads and for understanding which on-site behaviours actually correlate with conversion.

What is the free BigQuery export and why does it matter?

Google’s BigQuery export documentation confirms that every GA4 property can stream raw event data to Google BigQuery for free (within the BigQuery free-tier limits of 1 TB queries per month and 10 GB storage). This is one of GA4’s biggest functional advantages over UA, where the same feature was paywalled inside the $150,000+/year GA360.

What BigQuery export unlocks:

• Raw event data. Every event, every parameter, every user identifier. No sampling.
• SQL-level analysis. Joining GA4 data with CRM, ads, or product data.
• Long-term retention. Beyond GA4’s 14-month event limit.
• Custom attribution modelling. Calculate attribution your own way, not Google’s way.
• Integration with BI tools. Looker, Tableau, Power BI, Metabase all connect to BigQuery.
• Machine learning. BigQuery ML lets you build models on top of raw GA4 events.

BigQuery export is the most under-used free feature in GA4. Most small and mid-market sites have it disabled because nobody told them to enable it. Turning it on costs nothing, takes ten minutes, and produces compounding value: every event from this moment forward is queryable in SQL, joinable with other data, and retained beyond GA4’s UI limits.

Frequently asked questions

Is Universal Analytics still available in 2026?

No. Universal Analytics stopped processing new data on 1 July 2023, and Google removed all UA property access on 1 July 2024. Any site still using Google Analytics is now using GA4.

Is GA4 free?

Yes for most users. GA4 is free for properties under 10 million events per month. GA360 (the paid tier) is for high-volume properties with enterprise SLA, deeper sampling controls, and higher daily event quotas. BigQuery export is free within the BigQuery free-tier limits.

How long does GA4 keep data?

Event-level data: 2 months by default, configurable up to 14 months. Aggregated reports keep indefinitely. For longer retention, enable BigQuery export from day one and query historical data from BigQuery.

Does GA4 work without cookies?

Partially. GA4’s behavioural modelling fills gaps when consent is denied or cookies are unavailable. Real cookie-based attribution still produces the best data quality, but cookieless modelling means GA4 keeps producing usable reporting even as third-party cookies disappear.

How does GA4 compare to other analytics tools?

GA4 has the largest install base, free pricing, and tight Google Ads integration. Competitors like Plausible and Fathom emphasise simplicity and privacy. Mixpanel and Amplitude offer deeper product analytics for SaaS. Adobe Analytics serves enterprise. For most marketing-driven websites, GA4 is the default; the alternatives are worth considering when GA4’s complexity outweighs the integration benefits.

What this means in practice

GA4 has settled into its role as the default web analytics platform. The bumpy 2022 to 2023 transition is over, the platform has matured, and most of the original complaints (custom reporting limits, missing features, BigQuery friction) have been addressed in subsequent updates. The teams getting the most out of GA4 in 2026 are the ones that configured it properly, enabled BigQuery export, implemented Consent Mode v2, and treated GA4 reporting as one data source among several rather than the only source of truth.

For broader context, see our digital marketing strategy guide for how GA4 fits into measurement across channels.