What Is GA4? Google Analytics 4 Explained, Set Up and Actually Used

The most common GA4 mistake is one you can’t undo later: leaving data retention on its default. On a standard GA4 property, the user-level and event-level data behind explorations is kept for just two months unless you change it. Standard reports carry on working, so nobody notices until the day someone builds a funnel exploration comparing this festive season with last year’s, and the data simply isn’t there. Raising retention to fourteen months takes thirty seconds, but it only applies from that day forward.

That setting is a symptom of a bigger habit. GA4 gets installed, the tag fires, pageviews appear, and everyone assumes the job is done. It isn’t. Out of the box, GA4 doesn’t know which actions matter to your business, counts your own team as visitors, and often credits your payment gateway with your sales. The tool is capable. Most of the problems I find in client accounts come from treating it as something you install instead of something you configure.

What GA4 actually means

In plain terms, GA4 is the free tool from Google that tells you who visits your website or app, where they came from and what they did.

Formally, Google Analytics 4 is Google’s current analytics platform, which replaced Universal Analytics. It collects data from websites and apps through data streams, records every interaction as an event, and reports on those events by user, session and channel. A single GA4 property can combine a website and iOS and Android apps, so one customer’s journey across them can be seen in one place.

The nuance is in the event model. In older analytics, a “hit” might be a pageview, an event or a transaction, each treated differently. In GA4, everything is an event with parameters attached. A page view is an event called page_view with parameters like the page location. A purchase is an event called purchase with parameters like value, currency and items. That makes GA4 flexible, but it also means you have to decide which events matter and tell GA4 about them.

How GA4 works

Data flows into GA4 in layers:

• Automatically collected events: things GA4 records without extra setup, like first visits and session starts.
• Enhanced measurement: optional events you switch on in the data stream settings, including scrolls, outbound clicks, site search, video engagement on embedded YouTube players, file downloads and form interactions.
• Recommended events: events with names Google has defined, like add_to_cart, begin_checkout, purchase, sign_up and generate_lead. Using these names unlocks built-in reports, especially for e-commerce.
• Custom events: anything else you define yourself, such as a WhatsApp button click or a brochure download.

Each event can carry parameters. To use a custom parameter in reports, you register it as a custom dimension or metric in Admin. Until you do, it’s collected but mostly invisible.

Data sits at three scopes. User-scoped dimensions, like first user source, describe how someone first found you. Session-scoped dimensions, like session source / medium, describe how a particular visit began. Event-scoped dimensions describe a single action. Mixing scopes in one report is the most common reason GA4 numbers look odd.

GA4 also measures engagement. An engaged session lasts longer than ten seconds, includes a key event, or has at least two page views. The engagement rate follows from that:

Engagement rate = Engaged sessions ÷ Sessions

Bounce rate in GA4 is simply the share of sessions that weren’t engaged.

The settings to change in your first week

These are the settings I check on every GA4 property I audit, in roughly this order:

1. Data retention: raise event data retention to the maximum available for your property.
2. Internal traffic: define your office and team IP addresses as internal traffic, then activate the filter under data filters. Filters start in testing mode and do nothing until activated.
3. Unwanted referrals: add your payment gateways and any third-party checkout or booking domains to the list of unwanted referrals, so returning shoppers aren’t counted as new visits from the gateway.
4. Cross-domain measurement: if your journey crosses domains, like a main site and a separate booking domain, configure them together so one visit isn’t split in two.
5. Key events: mark the handful of events that represent real business outcomes as key events.
6. Product links: link Google Ads, Search Console and, if you use it, Merchant Center.
7. Custom dimensions: register the parameters you’ll want to report on.
8. Consent: if you need consent in your markets, set up consent mode so GA4 respects visitors’ choices.

Where to find things in GA4

• Reports, then Acquisition: user acquisition shows how new users first found you; traffic acquisition shows where each session came from. Use traffic acquisition for campaign reporting.
• Reports, then Engagement: events, pages and screens, and landing pages.
• Reports, then Monetisation: e-commerce purchases and item performance, if you send e-commerce events.
• Explore: free-form tables, funnel explorations, path explorations and segment comparisons for anything the standard reports can’t answer.
• Advertising: attribution paths and model comparison.
• Admin, then DebugView: live testing of events from your own device before you trust them.

A worked example: a travel agency’s missing enquiries

This is a hypothetical example with example numbers. Say a Kochi-based travel agency sells Kerala tour packages. Enquiries come through a website form and a WhatsApp button, and some customers pay a booking deposit through a payment gateway. GA4 was installed a year ago and never configured.

The owner’s report shows 9,000 sessions last month and 3 key events, which were purchases. The top converting source is the payment gateway’s domain. Organic search looks like it drives traffic but nothing else.

The fixes take one afternoon:

Problem	Fix	What changed
Form submissions not tracked	Created a generate_lead event on the thank-you page and marked it as a key event	140 enquiries now visible
WhatsApp clicks not tracked	Custom whatsapp_click event via Google Tag Manager, marked as a key event	210 WhatsApp clicks now visible
Payment gateway credited with sales	Added the gateway domain to unwanted referrals	Deposits credited to the original source
Staff traffic counted	Defined internal traffic and activated the filter	About 600 sessions removed
Retention at two months	Raised to fourteen months	Year-on-year explorations possible from now on

A month later, the picture is completely different. Organic search turns out to drive the largest share of form enquiries, mostly from blog posts about monsoon travel in Kerala. Paid search drives more WhatsApp clicks than forms. The owner shifts budget and content accordingly, based on data that had been invisible for a year.

GA4 vs Universal Analytics

	GA4	Universal Analytics
Data model	Everything is an event with parameters	Sessions and pageviews, with separate hit types
Web and app	Combined in one property	Separate properties
Main engagement metric	Engagement rate	Bounce rate
Default attribution	Data-driven	Last non-direct click
Goals	Key events	Goals, limited per view
Views	None; filters and data streams instead	Multiple views per property
Raw data export	BigQuery export available on standard properties	Only on the paid version
Status	Current	Retired

The shift that trips people up most is the loss of views. In Universal Analytics, teams used separate views for filtered data, test data and raw data. GA4 has no views, so filters apply to the whole property, and it’s worth testing them carefully before activating. The other big shift is mindset: GA4 expects you to decide what matters and configure it, where Universal Analytics came with more assumptions built in.

How to set up GA4 properly

1. Create the property and data streams. Add a web stream for your site and app streams if you have apps, and install the Google tag directly or through Google Tag Manager.
2. Change the first-week settings. Work through retention, internal traffic, unwanted referrals, cross-domain measurement and consent before you rely on any number.
3. Plan your events. Write down the actions that matter, use Google’s recommended event names where they fit, and create custom events only for what’s missing.
4. Test in DebugView. Trigger each event yourself and confirm it arrives once, with the right parameters, before marking it as a key event.
5. Tag your campaigns. Add UTM parameters to every external link you control, like emails, social posts and influencer links, so GA4 can credit them properly.
6. Link your Google products. Connect Google Ads and Search Console, and import key events into Google Ads if you’ll bid on them.
7. Build a few saved reports. Create the explorations your team will check every week, so people use GA4 instead of exporting screenshots.

Common GA4 mistakes

Expecting GA4 to match every other tool. GA4, Google Ads, Meta and your store all count differently, with different attribution rules, consent coverage and time zones. Chasing identical totals wastes weeks. Compare trends and proportions instead.

Marking too many events as key events. When scrolls, page views and button clicks are all key events, the reports can’t tell you what actually drives business. Keep key events for real outcomes.

Forgetting to register custom dimensions. Parameters you send but don’t register don’t appear in standard reports, which leads teams to believe tracking is broken when it isn’t.

Ignoring Unassigned and (not set). Growing Unassigned traffic usually means campaign links use nonstandard medium values. A lot of (not set) landing pages or sources often points to tagging or consent problems. Both are fixable once you look.

Judging yesterday’s numbers too early. GA4 can take a day or so to finish processing. Decisions made from same-day totals are often made on incomplete data.

How to tell if your GA4 setup is healthy

• Key events in GA4 move in line with real enquiries, sign-ups or orders in your own systems, even if totals differ.
• Your payment gateway and your own domain don’t appear as top referral sources.
• Unassigned and direct traffic stay small and don’t spike after campaigns.
• Explorations can reach back more than two months.
• Your team can answer “which channel drove last month’s leads?” from a saved report in a few minutes.

Key takeaways

• GA4 records every interaction as an event, and you decide which events matter by marking key events.
• Change data retention, internal traffic, unwanted referrals and cross-domain settings in the first week.
• Understand scopes: user, session and event dimensions answer different questions.
• Don’t expect GA4 to match ad platforms or your store exactly; compare trends instead.
• Tag campaign links consistently so GA4 can credit the channels that actually work.