What Is Google Analytics 4?

Google Analytics 4 (GA4) is Google’s current-generation web and app analytics platform, replacing Universal Analytics (UA), which permanently stopped processing data on July 1, 2023 for standard properties and July 1, 2024 for UA 360. Every new property created today is a GA4 property. Universal Analytics is fully decommissioned.

The defining shift in GA4 is its data model. Universal Analytics built its collection layer around session containers and distinct hit types: a pageview hit, an event hit, a transaction hit. GA4 replaces all hit types with a single primitive: the event. A pageview is an event named page_view. A session start is an event named session_start. A purchase is an event named purchase. Every event carries up to 25 custom parameters — key-value pairs providing dimensional context. This flat, extensible schema powers cross-platform reporting and a BigQuery export structure that is far easier to query than UA’s session-scoped schema. For the full breakdown of event types and parameters, see GA4 events and parameters.

GA4 launched officially on October 14, 2020. The UA sunset was announced in March 2022. Since July 1, 2023, any property still running on UA has been collecting no new data. Teams that did not export historical UA data before the read window closed have a clean break in their time-series. There is no automated path to port UA data into GA4 — the data models are incompatible. For anyone setting up analytics today, GA4 is the starting and ending point, not a migration target. The Google Analytics 4 glossary hub covers the broader vocabulary; this pillar covers the foundational model, setup, and the distinctions that cause the most operational confusion.

GA4 Key Terminology Reference

GA4 introduced new terms and redefined existing ones. The definitions below use GA4’s current vocabulary as of June 2026, including the 2024 rename of “Conversions” to “Key events” in the Analytics interface. Anyone referencing instructional content from before mid-2024 will encounter “Conversions” where the current interface shows “Key events.”

Key Takeaways

• GA4’s event-based model replaces UA’s session-and-hit-type model. Every interaction — including pageviews — is now an event with parameters. The data schema is flat, extensible, and BigQuery-native.
• Universal Analytics is fully decommissioned: standard properties shut down July 1, 2023; UA 360 shut down July 1, 2024. Historical data does not migrate automatically. Any new analytics implementation starts in GA4.
• Change data retention to 14 months on day one. The default is 2 months. The change is not retroactive. This is the single highest-impact day-one configuration step.
• In 2024, Google renamed GA4 “Conversions” to “Key events.” The word “Conversion” in the GA4 interface now refers exclusively to a Google Ads conversion action — a different object, counted differently.
• GA4 and Google Search Console measure different things. GA4 measures post-landing on-site behavior across all channels. GSC measures pre-landing search appearance for organic Google results only. Discrepancies between the two are expected and structurally inevitable.
• Consent Mode v2 is required for any new Google Ads audience or conversion connection on properties serving EEA users (enforced from March 2024). New properties must implement it before linking to Google Ads.

GA4 vs. Universal Analytics: Core Differences

Universal Analytics and Google Analytics 4 share a brand name and a measurement intent. The data models, session definitions, metric calculations, and storage schemas are fundamentally different. The table below covers the dimensions that cause the most operational confusion for teams migrating from UA documentation, UA goals, or UA-era reporting templates.

The bounce rate divergence is the most operationally significant metric change. In UA, a reader who arrives on a blog post, reads it completely, and leaves is a bounce. In GA4, that same session is an engaged session if it lasts 10+ seconds — and the engagement rate for quality content will be higher than UA’s non-bounce rate for the same traffic. Neither metric is wrong; they measure different things. Reporting templates that compare UA bounce rate directly to GA4 bounce rate are comparing incompatible metrics. For engagement metric depth, see GA4 engagement and user metrics.

The conversion counting divergence matters for paid media. UA goals counted once per session per goal. GA4 key events count multiple times per session if the event fires multiple times. When a GA4 key event is imported to Google Ads as a conversion, Ads counts per-click — which differs again from GA4’s per-event count. These differences appear as discrepancies in cross-platform reporting and are not data errors; they are counting-model differences. See GA4 conversions and key events for the full breakdown.

The GA4 Event-Based Model: Four Tiers of Data Collection

GA4 organizes all events into four tiers based on how they are collected and what setup they require. The tier determines whether an event fires with zero code, with a toggle, with developer work, or with developer work plus explicit parameter registration in the GA4 Admin.

The enhanced measurement tier deserves emphasis: scroll tracking (firing at the 90% scroll depth threshold), outbound click tracking, file download tracking, video engagement events for YouTube embeds, and form submission tracking all fire with zero code once the toggle is on. In Universal Analytics, every one of these required manual implementation. The Google Tag Manager integration makes recommended and custom event implementation accessible without direct code changes; GTM is the recommended implementation path for most sites that already use it for other tags.

For e-commerce implementations, recommended events follow a prescribed schema: purchase requires an items array with specific sub-parameters, and the currency parameter is required on all purchase events. Missing currency causes revenue to display as $0 in monetization reports even when purchase events fire correctly. For the complete e-commerce implementation reference, see GA4 ecommerce tracking.

GA4 Account, Property, and Data Stream Hierarchy

A GA4 property receives data from one or more data streams. One property handles an entire product — website, iOS app, Android app — in a single unified view. The table below covers the hierarchy from account to data stream, including the 360-only tiers that replace UA Views.

The property structure change from UA represents a significant architectural shift. In UA, a “View” was the primary access-control primitive — teams got filtered views so QA staff could see unfiltered data while analysts worked in filtered views. GA4 removes Views entirely for free properties. The replacement for access segmentation is the 360-only Subproperty. For free-tier properties, the recommended access-control pattern is a single property with carefully managed user permissions and data filters at the Admin level.

The additional configuration limits for free properties: 50 custom dimensions, 50 custom metrics, 100 audiences, 50 data streams per property, 2,000 properties per account. These limits are defined in Google’s configuration limits documentation. For BigQuery-heavy implementations and larger data architectures, see GA4 integrations and BigQuery.

Is Google Analytics 4 Free? Free Tier vs. Analytics 360

GA4 is free for the vast majority of use cases. The free tier includes the full standard reporting interface, Explorations, Google Ads integration, BigQuery export (up to 1 million events per day), predictive audiences, and predictive metrics. Analytics 360 — the paid enterprise tier, sold through Google Marketing Platform resellers — adds subproperties, roll-up properties, longer data retention, streaming BigQuery export, and elevated configuration limits.

The 2-month default data retention is the most consequential out-of-the-box limitation. Explorations — GA4’s custom analysis workspace — are subject to the retention limit. Any exploration spanning a date range older than the retention setting returns no data. For year-over-year analysis in Explorations, the 14-month maximum is the only option on the free tier, and it must be set on the first day the property is active. The change is not retroactive: data already outside the window is gone. Standard reports (the preset reports in the Reports section) are not subject to the same retention limit and retain aggregated data longer — but Explorations, which are the primary workspace for custom analysis and funnel building, are.

For most SMB and mid-market properties, the free tier is entirely sufficient. The threshold where Analytics 360 becomes cost-justified is typically above 25 million monthly events, or where the regulatory and SLA requirements of the enterprise tier are explicitly required by compliance. The correct action for any new property: change data retention to 14 months immediately, then evaluate 360 only if event volume or retention needs exceed the free tier.

GA4 Key Events vs. Google Ads Conversions: The 2024 Rename

In 2024, Google renamed what GA4 previously called “Conversions” — the in-Analytics concept of marking an event as business-critical — to Key events. Simultaneously, the word “Conversion” in the GA4 interface was reserved strictly for a Google Ads concept: a Google Ads conversion action, created from a GA4 key event but counted by Ads attribution rules, not GA4 rules.

The reason for the vocabulary split: before the rename, “conversions” appeared in both GA4 and Google Ads reports but counted differently. GA4 counted once per session per event (or multiple times if the event fired multiple times in a session); Google Ads counted per click. Cross-platform reviews routinely surfaced discrepancy that was not an error — it was a counting-model difference. The 2024 rename makes the distinction explicit in the interface: key event = the GA4-side concept; conversion = the Ads-side concept. The workflow is: (1) mark a business-critical event as a key event in GA4 → it appears in GA4 reports as a key event; (2) create a conversion in Google Ads from that key event → it appears in Ads reports, counted by Ads rules.

Any instructional content, screenshot, or walkthrough from before mid-2024 will show “Conversions” in the GA4 interface where current accounts show “Key events.” The underlying mechanic is identical; only the label changed. Marketers running GA4 properties set up before the rename will have seen the label change automatically in their interface — no data was affected. For paid media managers connecting GA4 to campaigns, see GA4 for PPC and lead generation.

How to Set Up a GA4 Property: Configuration Checklist

The 12-step sequence below covers creating a GA4 property from scratch and connecting it to the Google ecosystem. Steps 1–5 are the minimum for functional data collection. Steps 6–12 are required before running paid media or making data-driven decisions.

Steps 3 and 4 have outsized impact relative to their effort: enhanced measurement captures scroll, outbound, and form data immediately with one toggle; the 14-month retention change prevents a future data gap that would otherwise be permanent. MB Adv Agency runs these two steps as the first post-creation actions on any new property, before adding any custom implementation.

GA4 Reporting Identity: Blended, Observed, and Device-Based

Reporting identity controls how GA4 stitches cross-device user journeys in reports. The setting does not affect data collection — all events still fire regardless of the identity setting. It determines how the collected data is attributed to users across devices and sessions in reports and Explorations.

The modeling component in Blended identity is meaningful in EU/EEA markets. Consent decline rates in those markets range from 30–70% depending on implementation quality and jurisdiction. Without behavioral modeling, user counts in those markets are systematically deflated — not because users aren’t there, but because their data is excluded rather than estimated. Properties using Consent Mode v2 (required for new Google Ads connections from March 2024) enable the modeling layer automatically when Blended identity is selected. Privacy-safe data collection and Consent Mode v2 implementation are covered in depth in GA4 data collection and privacy.

MB Adv Agency sets new properties to Blended reporting identity as standard. The modeled user counts are closer to true user counts in consent-regulated markets than device-only counts, and the modeling methodology uses similar-user behavior from consenting users — not fabricated data. The reporting identity setting is in Admin → Reporting identity and can be changed at any time without affecting historical data collection.

GA4 vs. Google Search Console: Two Tools, Two Jobs

Google Analytics 4 and Google Search Console are not competing tools and are not redundant. They measure different things at different points in the user journey. The persistent confusion — that they should show the same organic traffic numbers — comes from the fact that both involve Google and both measure traffic. The measurement points are structurally different.

The numbers from these two tools will never match exactly, and that is expected. GSC records a click when a user clicks a Google Search result. GA4 records a session when its tracking tag fires on the landing page. Users who click a result but leave before the GA4 tag fires (fast bounces, ad blockers, page errors), users excluded by consent rules, and sessions where UTM parameters override the organic referrer all contribute to the divergence. The structural explanation: GSC measures Google-side behavior; GA4 measures website-side behavior. These are different events by definition.

The two tools combine most powerfully when linked. Linking GSC to GA4 (Admin → Property Settings → Search Console links) surfaces a Queries report inside GA4 that joins GSC query data with on-site behavior: you see which query drove the session AND what the user did after arriving. For the GSC side of this pairing, see what is Google Search Console, the GSC Performance report and search analytics, and the Google Search Console glossary hub. For Core Web Vitals data that crosses both tools, see Core Web Vitals and page experience.

Common GA4 Setup Mistakes and How to Fix Them

These are the configuration errors MB Adv Agency encounters most frequently in GA4 audits. Each has a specific, verifiable fix. Most are silent: they produce no error message in the interface and no warning in reports — the data simply degrades until someone runs an audit.

GA4 Report Surfaces: Standard Reports, Explorations, and BigQuery

GA4’s reporting interface organizes into four primary surfaces. Understanding which surface answers which question — and which limitations apply to each — is prerequisite to trusting the numbers.

Explorations and BigQuery serve different purposes. Explorations provide a no-code custom analysis workspace within the GA4 interface; they are the right tool for funnel analysis, path analysis, and user segment comparison. BigQuery export is the right tool for queries that join GA4 data with other data sources, that span more than 14 months, or that require SQL-level flexibility the Explorations UI does not provide. For the complete Explorations reference, see GA4 reports and explorations.

GA4 in 2025–2026: AI Assistant Channel, Ask Advisor, and Cost Data Import

Three platform additions in 2025–2026 are relevant to any current GA4 implementation. None require action to activate; all affect how data is classified and analyzed.

AI Assistant default channel group (May 13, 2026). Google added a dedicated “AI Assistant” channel to GA4’s Default Channel Group. Traffic from recognized AI chatbot referrers — ChatGPT, Gemini, and Claude are confirmed; Microsoft Copilot and Perplexity were added in August 2025 via custom regex guidance — is automatically classified under medium=ai-assistant. The limitation: AI assistant traffic arriving without a referrer header (in-app browsers, copy-paste navigation) still lands in Direct. Source: Search Engine Journal, May 2026.

Ask Advisor (Gemini-powered, beta). Announced at Google Marketing Live, May 2025. An agentic conversational interface in GA4 that answers natural-language queries against your property data: “Why did sessions drop last week?”, “Which pages have the highest key event rate?” As of June 2026, the feature remains in beta. It is available for English-language accounts with eligible configurations. Response quality is directly dependent on property configuration maturity — properties with poorly defined events or large data gaps produce less reliable Advisor answers.

Non-Google cost data import (2024–2025). GA4 added native automated integrations for cost, clicks, and impressions from Meta (Facebook + Instagram), TikTok, Pinterest (added September 26, 2025), Snapchat (September 17, 2025), and Reddit (July 21, 2025). This enables cross-channel CPA/ROAS comparison within GA4 without a third-party data connector. Only cost, clicks, and impressions are imported — not audience, creative, or placement data. For the full integrations reference, see GA4 integrations and BigQuery.

Frequently Asked Questions: Google Analytics 4

Data and methodology: Keyword volume and difficulty figures from Ahrefs, June 2026 (operator-supplied). UA shutdown dates from Google Analytics Help (support.google.com/analytics/answer/11583528). Event model and metric definitions from Google Analytics Help (answer/9322688, answer/11091422, answer/11986666, answer/12195621). Account structure and configuration limits from Google Analytics Help (answer/9679158, answer/12229528). Data retention from Google Analytics Help (answer/7667196). Analytics 360 features from Google Analytics Help (answer/11202874). BigQuery export limits from Google Analytics Help (answer/9823238). Reporting identity from Google Analytics Help (answer/10976610). Key events rename and conversions from Google Analytics Help (answer/13965727). Consent Mode v2 from Google Analytics Help (answer/14275483). Common mistakes from Analytics Mania (analyticsmania.com) and Search Engine Journal. AI Assistant channel from Search Engine Journal, May 2026. Ask Advisor from Google Analytics Help (answer/16675569). Cost data import dates from Mercado Global Media, 2025. Reviewed by MB Adv Agency, June 2026.