Google Analytics 4 (GA4) is Google's new generation website and application analysis platform, which belongs to the category. The event-centered data model replaces the traditional session model, unifies the user behavior tracking of websites and apps, has built-in predictive indicators (churn probability, purchase probability), intelligent insights and privacy security modeling capabilities, and supports in-depth analysis of raw data through BigQuery export.

Google Analytics 4

Core parameters and statistics

Google Analytics 4 (GA4) is Google's next-generation analytics platform for websites and apps. It replaces Universal Analytics' decade-old session model with an event model, decomposing user behavior into independent event streams instead of aggregated slices of page browsing sessions. The fundamental intention of this change is to unify the data calibers of websites and apps. Today, when cross-end user behavior has become the norm, the conversation model is naturally unable to accurately correlate the same user's continuous actions between the web and the app.

Projects	Public Information
Official positioning	Customer life cycle analysis platform to understand the user's entire journey across devices and platforms
Data model	Event-based, alternative session model (Session-based)
Core capabilities	Behavior analysis, conversion tracking, channel attribution, predictive indicators, intelligent insights
Built-in AI	Predictive probability (churn/purchase), intelligent insights (Proactive Insights), privacy and security modeling
Data export	BigQuery native export (supported by both standard version and 360)
Integrated Ecosystem	Google Ads, Display & Video 360, Search Ads 360, Google Cloud, Firebase, Salesforce Marketing Cloud, Google Ad Manager, AdMob, Google Play, Search Console
Supported platforms	Web, Android, iOS, API
Enterprise Edition	Analytics 360: Enhanced SLA, unsampled data, hourly refresh frequency
Free version core limit	10 million events per month (website + app combined), single hit size limit 128KB

Engineering implications of the event model: GA4 records each user interaction (page browsing, button click, scrolling, transaction) as an independent event, and each event has a timestamp, user ID and parameter set. This means that analytics are no longer bound by a "session start/end" framework, with the flexibility to define any sequence of actions as a transformation. The trade-off is that the event volume will be much higher than the traditional PV count - for high-traffic sites, the 10 million/month cap of the free version may become a hard constraint.

Supply conditions for predictive indicators: GA4's AI prediction functions (purchase probability, churn probability, revenue prediction) are not activated by default and require media resources to accumulate at least 28 days of historical event data and meet Google's model training threshold. Sites with small and medium traffic may not see forecast output for a long time.

User and market recognition

GA4 is the most widely deployed website analysis platform in the world, but behind its "extensiveness" there are two levels of market differentiation: passive migration of C-side small and medium-sized sites and active adoption by B-side enterprises.

Market Coverage: Since Google stopped processing data with Universal Analytics in July 2023, almost all sites using Google Analytics have been passively or actively migrated to GA4. The penetration rate of Google Analytics among the top 100,000 websites in the world has long ranked first, but this number reflects more of UA’s historical accumulation rather than users’ active choice of GA4.

Quantitative signals of enterprise cases: Google's official case page disclosed several verifiable customer results - 412 Food Rescue shortened the reporting time by 50% ("cut our reporting time by 50%"); Lider achieved an 85% reduction in single customer acquisition cost and an 18X conversion rate improvement through GA4 ("85% decrease in CPA and 18X conversion rate improvement"); McDonald's Hong Kong used GA4 to achieve in-app order growth 550%. These cases illustrate that GA4's attribution and audience capabilities have quantifiable output in large traffic scenarios, but the specific results are highly dependent on the company's own data foundation and analysis maturity.

Community and third-party ecosystem: GA4 has spawned a huge third-party tool ecosystem (data dashboard, audit tool UTM builder GA4 to BigQuery pipeline tool), and its API and BigQuery export are the entrance to independent data products and consulting services. The labor market for professionals certified in Google Analytics continues to grow around GA4, which itself is a manifestation of the platform lock-in effect.

Prerequisites: The upper limit of the value of GA4 depends on the quality of the media resources. If the team is still stuck with simple page view tracking and does not deploy custom events, conversions and user IDs, GA4's cross-platform, prediction and attribution capabilities are basically idle.

Cost advantage

The cost structure of GA4 presents a clear two-tier model of "free standard version + enterprise version 360", but the hidden costs and scale bottlenecks are worth dismantling separately.

C-side/Personal Webmaster: The standard version is free to use, and the monthly event quota of 10 million is enough for most small and medium-sized sites. For payment plans that exceed the quota, please refer to the official real-time page. The real cost for individual users lies in the learning curve - switching from the conversational thinking of UA to the event thinking of GA4, and the ongoing adaptation cost caused by Google's frequent adjustments to the reporting interface.

Developer/Data Analyst: Raw event data can be obtained through BigQuery export for in-depth modeling. BigQuery is billed based on the amount of data scanned (the Web UI has a free monthly quota), and the cost is controllable for high-frequency query scenarios. GA4's Data API v1 supports programmatic report pulling and is suitable for embedding in internal systems. API calls are managed according to project quotas and there is no additional billing.

Enterprise/medium and large sites: Analytics 360 is for enterprises whose monthly event volume exceeds the free version or needs SLA guarantee. The price needs to be confirmed with the sales team. Core benefits of the 360 ​​version include: hourly data refresh (daily for the Standard version), unsampled data reporting, enhanced credits (such as higher event/user/parameter caps), sub-property and summary reporting. Before purchasing, enterprises need to focus on verifying SLA terms, data residency options, and price locking mechanisms during the contract period.

Hidden costs: For sites with a monthly event volume approaching tens of millions, the free version limit may pose actual pressure. In addition, the event schema design of GA4 determines the flexibility of downstream analysis - if the initial burying point is too extensive, the cost of subsequent retroactive transformation will be much higher than that of one-time planning. The increase in scan volume in high-frequency BigQuery export scenarios also needs to be factored into the operating budget.

Main functions

The functional system of GA4 is designed around "full life cycle user analysis", covering everything from collection to insight. The core modules are as follows:

• Event model and custom events: Record each interaction as an independent event (page_view, scroll, click, purchase, etc.), and support custom event names and parameters. Applicable tasks: Any user behavior can be used as an independent analysis dimension, no longer limited to page browsing paths.
• Predictive Metrics (AI Powered): User-level predictions based on machine learning, including purchase probability (probability of a transaction within 7 days), churn probability (probability of not returning within 7 days), revenue forecast (expected revenue in the next 28 days). Applicable tasks: Build high-value user lists, trigger proactive marketing interventions, and optimize advertising audiences.
• Intelligent Insights (Proactive Insights): Automatically identify abnormal changes in the data (such as a sudden drop in traffic in a certain channel, abnormal changes in the conversion rate of a certain type of user), and push it to the top of the report in a natural language summary. Applicable tasks: Help the operations team quickly locate data changes that require attention and reduce coverage blind spots in manual monitoring.
• Explorations: Provides 8 analysis templates including free format, funnel analysis, path analysis, segmentation overlap, same group analysis, user exploration, and user life cycle. Applicable tasks: Flexible multi-dimensional data exploration for analysts and advanced users, independent of preset reports.
• Advertising workspace and attribution modeling: Check the delivery performance and conversion contribution of each channel, and support multiple modeling methods such as data-driven attribution (Data-Driven Attribution). Applicable tasks: Evaluate the efficiency of marketing budget allocation and optimize multi-channel portfolio solutions.
• BigQuery native export: Automatically export GA4 raw event data to BigQuery, supporting SQL in-depth query and external model training data construction. Applicable tasks: Data teams perform custom modeling, business reporting and data mining outside of the analytics platform.
• Audience Generation and Activation: Based on event bars

Create audiences based on dimensions such as files, user attributes, and prediction scores, and synchronize them to advertising platforms such as Google Ads, Display & Video 360, etc. with one click. Applicable tasks: Translate analytical insights directly into ad targeting strategies to shorten the distance from discovery to action.

Functional synergy: The value of predictive indicators is not only to tell you "which users are likely to churn", but also to directly output the audience list and synchronize it to Google Ads - which means that the process from "analyzing high-risk users" to "placing retention ads on these users" can be completed within the same platform, without the need for data pipeline transfer.

Model and version evolution

The version evolution of GA4 is actually a switch between generations of Google Analytics products, rather than an increment of software version numbers in the traditional sense.

First generation: Classic Analytics (Urchin era)

Launched after Google acquired Urchin Software in 2005, it uses page tagging technology, and the report is based on page views and visits. This generation laid the foundation for the free Google Analytics tool, but the data model looks crude today.

Second generation: Universal Analytics (session model era)

Released in 2012, it adopts the three-layer model of "visitor-session-browsing/event" and introduces capabilities such as audience analysis, target conversion, and enhanced e-commerce. UA has become the de facto standard for website analysis worldwide, but its session model is naturally unable to unify website and app data - the same user's cross-end behavior will be recorded as two independent sessions and cannot be correlated.

The third generation: Google Analytics 4 (event model era)

Released as beta in October 2020, fully replacing UA in July 2023 (UA standard media resources stop processing data). Core changes in GA4 include:

• The event model replaces the session model and unifies web and app data
• Built-in machine learning predictive capabilities
• User-level data retention (default 2 months, adjustable to 14 months)
• BigQuery export available with free version
• Anonymous IP by default, full IP address is no longer stored

Remaining issues in migration: The interface and data caliber of GA4 are significantly different from those of UA, and many historical reports accumulated during the UA era cannot be directly transplanted. Google provides a migration guidance tool and comparison report from UA to GA4, but historical data needs to be restored from UA backup or BigQuery and cannot be automatically inherited to GA4.

Technical advantages

The technical advantages of GA4 are not reflected in a single performance indicator, but in the flexibility of the data model and the depth of collaboration in the Google ecosystem.

Engineering bonus of the event model: GA4 turns each user behavior into an event record with parameters, allowing for flexible burying without a fixed Schema. Compared with UA's "page = one PV in a session" fixed frame, GA4 can record three-dimensional data points such as "the user clicked the video play button + video ID + current progress bit at the 30th second" - this is particularly critical in a cross-platform scenario. "Sliding browsing" on the iOS side and "page browsing" on the Web side can ultimately be mapped to "content browsing" events with the same parameter structure.

Embedding of machine learning capabilities: GA4’s prediction model does not require users to manually train, but is a general prediction model built by Google based on aggregated data at the top level. This means that users only need to meet the basic data threshold to obtain user-level scores such as "purchase probability 0.86". But this also means that the model has limited transparency - users cannot know which features drive predictions, cannot adjust model parameters, and the prediction output is a black box that is "out of the box but not customizable".

Privacy-Safe Modeling: In the context of the gradual withdrawal of iOS ATT (App Tracking Transparency) and third-party cookies, GA4 introduces behavioral modeling to fill the gap in observation data. Its core logic is to model and estimate user behavior based on observable user patterns that cannot be directly tracked due to privacy restrictions. This mechanism is manifested as a mixture of "modeling events" and "observation events" at the report level. However, Google has not disclosed the modeling accuracy verification method. Users need to evaluate their trust in the modeling data by themselves.

BigQuery's deep integration: Writing raw event logs directly to BigQuery is the technical barrier that distinguishes GA4 from almost all competing products - the advertising platform puts click/impression logs and analysis event logs in the same data warehouse by default. Analysts can use SQL to complete the full-link query "from ad exposure to on-site conversion" without manually splicing multiple data sources.

How to use

The implementation path of GA4 is divided into data collection layer, configuration layer and analysis layer, and different roles intervene from different entrances.

How to use	Suitable for the role	Key actions	Cost
Web console	Operations, marketing, analysis team	Create media resources, configure events and conversions, view reports	Free
Google Tracking Code (gtag.js)	Front-end Developer	Single-page deployment of baseline code, extended with custom events	Free
Firebase SDK	App Developers	Access iOS / Android analytics events	Free
Google Tag Manager (GTM)	Tag Management Team	Manage all tracking tags through GTM containers without modifying site source code	Free
Data API v1	Data Engineer	Programmatically pull report data and integrate into internal system	By API quota
BigQuery export	Data team	Export raw event data in real time for SQL in-depth analysis	BigQuery scan billing
Analytics 360	Enterprise	Obtain enhanced quota SLA and sub-media resources after signing the contract	Business confirmation required

Typical implementation steps:

1. Create GA4 media resources in the management panel and obtain the Measurement ID (G-XXXXXXX).
2. The website uses gtag.js or GTM to deploy baseline code; the app integrates Firebase SDK.
3. Configure custom events and key conversion events (such as purchase, sign_up).
4. Set user attributes and audience conditions.
5. Verify whether the data flow is stable (if the real-time report is visible, it means the collection is normal).
6. Configure a BigQuery connection (optional) and start raw data export.
7. Enter the Discovery module or Standard Report to begin analysis.

For new sites, the time from deploying baseline code to usable data appearing in reports is typically 1-2 hours; historical data import (UA to GA4) requires additional configuration. Google also provides free Analytics Academy and Skillshop courses, which are suitable for teams to systematically learn the configuration and analysis process of GA4.

Product Pricing

The pricing structure of GA4 continues Google's consistent "free standard version + enterprise version on-demand pricing" strategy.

Standard Edition (Free):

• Media resource quota: up to 200 media resources per account
• Event collection volume: 10 million events per month (website + app combined)
• Data retention: User-level data defaults to 2 months, and can be set to a maximum of 14 months
• Report function: full standard report + exploration and analysis 200 templates/user
• BigQuery export: supports native free export
• Measurement Protocol: supports server-side event sending

Analytics 360 (Enterprise Edition):

• Event collection: higher monthly quota (subject to contract)
• Data refresh: once an hour (once a day for standard version)
• Report Sampling: No Sampling Report
• Sub-media resources and summary reports: unified management of multiple brands and multiple channels
• SLA: Availability and Data Integrity Commitment
• Data residency: supports regional data storage selection
• Enhanced custom dimension/metric restrictions

Actual prices require contacting Google sales team for a quote. When evaluating 360 Edition, enterprises should focus on aligning the peak monthly event volume with 360 Edition's quota coverage to avoid overage charges.

Application scenarios

The implementation scenario of GA4 focuses on data-driven marketing and product operations teams, covering the complete link from customer acquisition to retention:

• E-commerce and retail funnel analysis: Track the conversion rate of each section from ad clicks → product browsing → additional purchases → payment, combine predictive indicators to identify user groups with "high purchase probability but have not yet placed an order", and export the audience to Google Ads for precise follow-up investment. Key points of verification: Whether the event schema completely covers key purchasing actions, and the impact of attribution model selection on channel distribution results.
• Content media and user retention: Analyze article/video reading depth, interactive behavior (comments, collections, sharing) and frequency of return visits, use churn probability prediction scores to lock in low active users in advance, and conduct recall intervention through push or email. Key points to verify: The correct definition of content interaction events (for example, a video playback progress of 75% is considered a "valid view"), and the cross-device association coverage of user IDs.
• B2B and Lead Gen: Track lead-type conversions such as form filling, demo reservations, and data downloads, and combine UTM parameters and channel attribution to analyze the lead quality and final transaction conversions of each marketing channel. Key points to verify: The cross-session path of leads from "submit form" to "transaction" needs to be customized in BigQuery. GA4 preset reports have limited support for B2B long-cycle conversions.
• App behavior and monetization analysis (Firebase integration): For products that operate both the web and the app, GA4 provides the ability to uniformly collect data from both ends through the Firebase SDK - the user's complete path from web browsing to app purchase can be reconstructed. Key points to verify: Schema consistency between App events and Web events in the same media resource.

Applicable people

GA4's multi-layered capability system enables it to serve four core roles:

• Operations and Marketing Team: Utilize standard reports, channel attribution and audience functions to complete daily traffic monitoring, activity effect evaluation and ad targeting optimization. You need to at least understand the event model and conversion configuration. Typical task cycles range from hourly (real-time monitoring) to monthly (attribution review).
• Data Analysts & BI Teams: Perform flexible data mining with Discovery Analytics’ 8 analysis templates, or build custom attribution models, user lifecycle curves in BigQuery. Depends on SQL proficiency and familiarity with GA4 Schema.
• Product Manager and Growth Team: Track feature adoption rate (Feature Adoption), user retention curve and conversion funnel based on event model to support product iteration decisions. You need to master the design of user attributes and event schemas, otherwise "the desired analysis dimensions cannot be found in the data."
• Google Advertising Team: Sync GA4 audiences directly to Google Ads, Display & Video 360 for precise targeting based on behavior and predictive scores. Audience quality is directly affected by the data completeness of your GA4 properties.

Does not fit boundaries:

• Minimalist sites that do not require multi-dimensional, event-level analysis (such as pure blogs, corporate presentation sites) - the complexity and learning cost of GA4 far exceed its needs.
• Enterprises with strict control requirements for data sovereignty - GA4 data is stored in Google data centers. Although the 360 ​​version provides regional options, the overall data governance flexibility is not as good as self-built infrastructure.
• Businesses that focus on deep long-term attribution (such as non-linear B2B conversions over 90 days) - GA4's preset attribution window is 30-90 days, and more complex attribution models need to be built by yourself through BigQuery.

Summary and Outlook

The core competitiveness of Google Analytics 4 is that it locks the AI ​​prediction capabilities of the free-level analysis platform and the Google advertising ecosystem into the same data layer: user behavior data is both the raw material for analysis and the input signal for advertising targeting - this relationship has become the moat of GA4 under the increasingly stringent regulatory environment of data isolation. Its event model is the right direction at the architecture level compared to UA, clearing structural obstacles for cross-platform analysis.

The current main limitations include: the event quota of the free version is a hard bottleneck for medium and large sites; the AI ​​prediction model is a black box output and cannot be adjusted to participate in audits; the separation of the interface and UA brings continuous migration and learning costs; the accuracy verification method of privacy modeling is not disclosed, and users need to judge their own trust in the modeling data.

Procurement and Adoption Risk Assessment: Starting from GA4 is the lowest risk path for new sites or teams; teams still using UA are recommended to accelerate migration after completing data backup, but at least 2-4 weeks of staff training and Schema design time need to be reserved. For enterprises with monthly event volume approaching tens of millions and requiring SLA or data residency guarantees, quota caps, overage rates, and contract lock-in terms should be aligned before signing the 360 ​​version contract. It is not recommended to use GA4 as the only data infrastructure - in business-critical attribution and prediction decisions, it is recommended to include the raw data exported by BigQuery into an independent audit link to establish the ability to cross-validate the GA4 output results.

Version Info

• Google Analytics 4 ：Unified website and app analysis based on event models, providing cross-platform user paths, predictive indicators, intelligent insights and privacy security modeling; the specific capability boundaries are subject to the latest official page. （2020-10-14）
• Universal Analytics ：The previous generation of analytics based on the session model stopped processing standard media resource data in July 2023 and was fully replaced by GA4. （2012-10-01）
• Classic Analytics (Urchin) ：The predecessor of Google Analytics, derived from Urchin Software acquired in 2005, uses page tags and simple reporting models. There is no official precise date yet. （2005-11-01）