What Is Google Analytics? How GA4 Works, Explained

Google Analytics is Google's measurement tool for websites and apps. It tells you who visited, which channel brought them, what they did on the page, and whether the visit ended in a sale or an enquiry. The version in use today is Google Analytics 4, or GA4, which is free for ordinary businesses, while large organisations that need higher limits can buy Analytics 360.

This article explains how GA4 collects data, how it differs from Universal Analytics, what an account and a property are, which reports a business owner should open every week, and how the Google Ads, Search Console and Consent Mode connections work. Every limit quoted here comes from the Google Analytics Help Center as of October 2026.

What Google Analytics is and what it measures

Put as briefly as possible, Google Analytics is a system that receives a "signal" from a website or app each time a user does something, then organises those signals into reports. On websites the signals are sent by tracking code called the Google tag (gtag.js) or through Google Tag Manager; on iOS and Android apps they are sent through the Firebase SDK.

The questions GA4 can answer fall into a few broad groups:

• Where users came from, such as organic Google search, Google Ads, social media, email, or typing the URL directly
• What users did on the site, such as which pages they opened, how far they scrolled, outbound link clicks, file downloads, or starting a form
• Business outcomes, such as orders, revenue, leads from form submissions, or clicks on a LINE button that you define as important
• User characteristics, such as device, browser, country, city and language, plus age and interest data when Google signals is switched on

What Google Analytics does not do is tell you which words people searched on Google before arriving. That query data comes from Search Console, which has to be linked separately. GA4 is also not a CRM that stores customer names or phone numbers. Google's policies prohibit sending personally identifiable information, such as email addresses or phone numbers, into Analytics.

How GA4 collects data: the event-based model

The core idea in GA4 is that everything is an "event". Opening a page is an event named page_view, scrolling is an event named scroll, starting a session is an event named session_start, and buying a product is an event named purchase. Each event carries details with it in the form of "parameters". A purchase event, for example, carries the order value, the currency and the list of items.

Where events come from

Events in GA4 have four sources. The first is events collected automatically as soon as the tag is installed, such as first_visit, session_start and user_engagement. The second is Enhanced measurement, which you switch on in the web data stream settings without writing extra code. It covers page_view, scroll, click (outbound links to other domains), view_search_results, file_download, form_start, form_submit, and video_start, video_progress and video_complete for embedded YouTube videos.

The third is recommended events, where Google has already defined the names and parameters, such as generate_lead, sign_up, add_to_cart, begin_checkout and purchase. Using Google's names means some built-in reports, such as the Ecommerce purchases report, fill themselves in. The fourth is custom events for behaviour specific to your business, such as clicking an "add us on LINE" button or opening a branch map.

Parameters, custom dimensions and key events

A single event can send up to 25 parameters on the standard version (100 on 360). Custom parameters do not appear in reports until you register them as a custom dimension or custom metric in Admin. The standard version allows 50 event-scoped custom dimensions and 50 custom metrics per property. A common mistake is sending a parameter but forgetting to register it, which leaves the period before registration missing from standard reports.

Events that matter to the business, such as a form submission or a purchase, should be marked as a "Key event". This is the name Google has used in GA4 since 2024 in place of "conversion". The word conversion now refers to a key event that is used for measurement and bidding in Google Ads. The standard version allows 30 key events per property and 360 allows 50.

How GA4 differs from Universal Analytics

Universal Analytics (UA) was the previous version. It used a session-based model with separate "hit" types such as pageview, event and transaction. GA4 folds all of these into one event format, which is what lets web and app data live in a single property.

According to Google's help pages, standard UA properties stopped processing new data on 1 July 2023. UA 360 properties with a current order received an extension to 1 July 2024, and Google began removing the UA interface and API from the week of 1 July 2024. In practice, UA data that was not exported before then can no longer be opened, and old data was never moved into GA4 automatically.

A business owner will notice several differences straight away. First, bounce rate has been redefined in GA4. A session counts as an "engaged session" when it lasts longer than 10 seconds, includes a key event, or has 2 or more page views. Engagement rate is the share of sessions that were engaged, and bounce rate is the remainder, so the two always add up to 100%. Second, GA4 has no UA-style "views"; filtering is done with data filters and property-level settings instead. Third, user and session counts in UA and GA4 normally do not match because the counting definitions differ, so the two should not be joined on one chart without a note.

What an account, a property and a data stream are

Google Analytics has three levels. The top level is the Account, which usually represents one company; an account can hold up to 2,000 properties. The middle level is the Property, where the data and reports actually live. Time zone, currency, data retention and the Google Ads links are all set at this level. The bottom level is the Data stream, the source that sends data into the property. There are three types: Web, iOS app and Android app.

An example structure (hypothetical, for illustration): an online shop with Thai and English pages on the same domain, plus iOS and Android apps, should use one property with one web stream and two app streams, then separate languages with a dimension such as the URL path or page language rather than building one property per language. Splitting properties means a single user who switches language is counted as two people.

Each web stream has a Measurement ID that starts with G-, which is the value you put in the tag or in Google Tag Manager. If a site still carries old code that starts with UA-, that code no longer sends data anywhere and can be removed.

Reports a business owner should read

GA4 has a lot of reports, but if you only have a few minutes a week, these answer the most business questions.

Realtime

Shows users and events from the last 30 minutes. Use it to check that the tag still fires after a site change, or to confirm that a campaign you just launched is actually sending people. Do not use it to judge performance, because the data has not been fully processed yet.

Acquisition: User acquisition and Traffic acquisition

These two look alike but answer different questions. User acquisition groups new users by the channel that first brought them (first user source), so it tells you which channels find new customers. Traffic acquisition groups each session by its own channel, so it tells you which channels bring people back. If ads pull in many new visitors but sales happen in later sessions that come back through brand search, you only see that picture when you read the two reports side by side.

Engagement: Pages and screens, and Events

Shows which pages get the most views and how long people stay, and which events fire most often. Use it to find pages with heavy traffic but low engagement, which are often pages whose content does not match what people expected from the search result or the ad.

Key events and Monetization

If key events are set up correctly, these reports answer how much the site actually produces. An online shop sees revenue, order count and best-selling products, while a service business sees lead counts by channel.

Explore

The space for building your own reports, such as a Funnel exploration to see at which checkout step users drop off, or a Path exploration to see the sequence of pages people move through. Two things to know: Explorations can only show data inside the retention period you set, and on the standard version they are sampled when a query exceeds 10 million events.

Is Google Analytics free, and what does GA4 360 add?

The standard version of GA4 costs nothing and is enough for most businesses. Analytics 360 is the paid version for organisations that need higher limits, longer retention and a service-level agreement (SLA). Google does not publish a fixed price on its help pages; pricing comes from Google's sales team or an authorised reseller. The table below summarises the differences that matter in daily use, with figures taken from the GA4 limits page in the Google Analytics Help Center.

The data retention setting

Many people assume that retention means all reports disappear after 2 or 14 months. In fact the setting applies to user-level and event-level data tied to cookies or User-ID, which is what Explorations and funnel reports use. Standard aggregated reports are not affected. Age, gender and interest data always use a 2-month period.

Many new properties are still sitting on 2 months, the shortest option. If you want to compare behaviour year over year in Explorations, go to Admin and change it to 14 months. Lengthening the period applies to data that has not yet been deleted. When you shorten it, data outside the new window is removed in the next monthly deletion run. Changes take effect after a 24-hour wait, and you can revert within that window without losing data.

Connecting to Google Ads and Search Console

Google Ads

Linking GA4 to Google Ads requires the Editor or Administrator role on the GA4 property and administrative access to the Google Ads account. Once linked, you can turn GA4 key events into Google Ads conversions for measurement and bidding, use GA4 audiences for remarketing (Google signals or user-provided data collection must be active), and see Google Ads clicks, cost and CPC inside GA4 within about 48 hours. A property can have up to 400 links, and a link to a manager account (MCC) counts as one.

Two things to check: auto-tagging in Google Ads must be on so ad clicks are attributed correctly, and conversion counts in Google Ads and key event counts in GA4 rarely match, because the two systems date conversions and assign credit differently.

Search Console

Linking Search Console requires the Editor role in GA4 and verified ownership of the Search Console property. One web stream can link to only one Search Console property. After linking you get two reports: Queries, which shows organic search terms, and Google organic search traffic, which shows landing pages alongside GA4 metrics. This report collection is unpublished by default, so you have to open Library and publish it yourself. The data arrives about 48 hours after Search Console collects it and goes back no more than 16 months, the period Search Console keeps.

Privacy and Consent Mode

Consent Mode is the mechanism that tells Google's tags which kinds of data collection a user has agreed to. It uses four main consent states: analytics_storage (analytics cookies), ad_storage (advertising cookies), ad_user_data (sending user data to Google for advertising) and ad_personalization (personalised advertising).

There are two ways to implement it. Basic mode blocks the tags until the user answers the cookie banner; if the user declines, nothing at all is sent to Google. Advanced mode loads the tags when the page opens; while consent is not given, the tags send cookieless signals, which lets Google run behavioural modelling to estimate the behaviour of users who declined.

Behavioural modelling in GA4 has minimum thresholds. The property needs at least 1,000 events per day with analytics_storage set to denied for at least 7 days, and at least 1,000 daily users with consent granted on at least 7 of the previous 28 days. Many small sites do not qualify, so their numbers simply fall in line with the share of people who decline cookies.

What this means for businesses in the Thai market

Thai websites should check three things in particular. The first is the Personal Data Protection Act (PDPA), which has led many Thai sites to add cookie banners. If a banner blocks the tags without passing consent states through Consent Mode, GA4 numbers drop with no explanation and Google Ads receives fewer conversion signals. Have your own legal team confirm the right consent approach; this article is not legal advice.

The second is contact through LINE and phone. Many Thai businesses close the sale in chat. If you do not set up events for LINE button clicks or call buttons, GA4 will miss most of what the site produces, and Google Ads will bid without its most important signal.

The third is the property's time zone and currency. Set them to Asia/Bangkok and THB to match the Google Ads account so daily totals in both systems can be compared. And if the site uses Thai-language URLs, the Pages and screens report may show paths encoded as %E0%B8..., which are hard to read; adding the Page title dimension makes the report much easier to scan.

Limits of Google Analytics worth knowing

GA4 does not count everyone who visits. Users with ad blockers, or who decline cookies under Basic consent mode, are not counted. Some reports may hide rows with few users when Google signals is on (data thresholding) to prevent identification of individuals, and Explorations on the standard version are sampled above the threshold. GA4 numbers are therefore best used for trends and proportions rather than as an accounting ledger that must match the back office to the last baht. For actual sales, the store platform or ERP remains the main source of truth.

GA4 also does not backfill. Data starts from the day the tag starts working, so an event you forgot to set up three months ago will never have history. It pays to plan events completely at the start, and to export to BigQuery if you need raw data kept longer than GA4's retention period.

Frequently asked questions about Google Analytics

What is the difference between Google Analytics and GA4?

GA4 is the current version of Google Analytics, not a separate product. When people say Google Analytics today they mean GA4, because the older Universal Analytics stopped processing data in 2023 and its interface was shut down from July 2024.

Is Google Analytics really free?

Yes, the standard version of GA4 costs nothing and has no limit on the number of distinct event names on web streams. Costs only arise if you buy Analytics 360 or use BigQuery beyond Google Cloud's free quota.

Do I need Google Tag Manager to use GA4?

No, GA4 can be installed with the Google tag directly. Google Tag Manager makes it easier to add or change events without editing site code every time, which suits sites that need to track LINE buttons, forms or tags from several ad platforms.

Why don't GA4 numbers match Google Ads or my back office?

Because each system counts differently. Google Ads credits a conversion to the date of the ad click, GA4 records it on the date the event happened, and GA4 cannot see users who decline cookies or use blockers. Some gap is normal, but an unusually large one is a reason to check the tag setup and Consent Mode.

Getting reliable data out of Google Analytics

Installing the GA4 tag takes minutes, but the data only becomes useful when the property structure, key events, Consent Mode and Google Ads link are set up properly from the start. If you want specialists to audit or rebuild your setup, see the Google Analytics 4 setup and migration service. If your data is spread across GA4, your store platform and a CRM, the data integration and consolidation team can bring it into one dataset. Once the data is in order, decisions on Google Ads and SEO can rest on real numbers.