GA4 confuses people because it threw out the vocabulary everyone had learned. Once you understand that everything is an event, the interface stops fighting you. Thirty minutes to a setup you trust and a routine you'll keep.
⏱ 4 min read
Google Analytics 4 (GA4) records what people do on your site: which pages they view, what they click, how long they stay, whether they buy or enquire. It is free, it is the only version still running — the previous generation, Universal Analytics, stopped processing data on 1 July 2023 — and it works across websites and apps in one property.
If you used the old Analytics, the frustration is real and it has one root cause. Universal Analytics was session-based: it counted visits and grouped pageviews inside them. GA4 is event-based: every single interaction is an event with a name and some parameters. A pageview is an event called page_view. A scroll is an event. A purchase is an event. Sessions still exist, but they are assembled from events rather than the other way around.
One model for web and app. Interactions that used to need custom code — scroll depth, outbound clicks, file downloads, video engagement, site search — tracked automatically. And a reporting system where you decide which events matter rather than working within fixed categories.
Familiarity, mostly. Bounce rate is no longer the headline metric. Goals are now "key events". Reports look thinner out of the box because GA4 expects you to configure and customise. And the data is sampled and modelled in places, so GA4 and Search Console will never agree exactly — they count different things, at different moments, with different rules.
Search Console ends at the click; GA4 starts there. Search Console says a person searched for something and clicked your page. GA4 says they then read two more pages, spent ninety seconds, and submitted a form. Neither tool answers the other's question, and linking them (Admin › Product links) puts a limited slice of Search Console data inside GA4.
Welcome to the Google Analytics Tutorials — Google Analytics. Google's own orientation to the GA4 tutorial series — a good two-minute framing before you start clicking. Open on YouTube ↗
GA4 was built in a period of tightening privacy rules, and it shows. Cookie consent banners mean some visitors are never measured. Ad blockers stop the tag loading for a meaningful slice of technical audiences. Google fills some of the resulting gaps with modelled data rather than observed data, which is why numbers can shift slightly after the fact.
The practical consequence is not that GA4 is unreliable — it is that GA4 measures a consistent sample rather than a census. Trends are trustworthy; absolute totals are an undercount of unknown size. If you work with EU visitors, Consent Mode changes how the tag behaves before and after consent, and is worth configuring properly rather than bolting on later.
| Universal Analytics | GA4 equivalent |
|---|---|
| Goals | Key events |
| Bounce rate (headline metric) | Engagement rate |
| Views (within a property) | Properties and comparisons |
| Behaviour reports | Engagement reports |
| Custom reports | Explorations |
⏱ 4 min read
GA4 has three nested levels. An account holds one or more properties; a property holds one or more data streams. A property is the reporting unit — normally one per business. A data stream is one source feeding it — your website, an iOS app, an Android app.
G-XXXXXXXXXX.<head>. Google Tag Manager is the cleaner long-term option if you will add more tracking later.page_view and session_start arrive. If nothing appears, the tag is not firing — fix that before anything else.| Setting | Where | Why it matters |
|---|---|---|
| Data retention | Admin › Data settings | Defaults to 2 months. Change to 14. It is not retroactive. |
| Internal traffic filter | Admin › Data streams › Configure tag settings | Keeps your own visits out of the numbers. |
| Unwanted referrals | Same menu | Stops payment gateways showing up as traffic sources. |
| Google Search Console link | Admin › Product links | Brings query data into GA4's reports. |
Data retention is the one that stings. Leave it at two months and in March you will not be able to build an exploration covering January. Change it during setup, on every property, every time.
Pasting the gtag snippet into your theme works and takes two minutes. Google Tag Manager takes twenty and pays for itself the first time you need a third tag. GTM gives you one container to manage every script — GA4, ads pixels, chat widgets — with a preview mode and version history, so a broken change is one click from being rolled back. For a site you will maintain over years, or any site where multiple people add tracking, GTM is the right default. For a single-page brochure site that will never change, the direct snippet is fine.
"The plugin says it's connected" is not verification. Load the live site in an incognito window, then check Realtime for your own visit. For more detail, enable debug mode and watch DebugView, which shows each event with its parameters as it arrives — the only way to confirm a custom event carries the values you think it does.
Two failure modes are worth checking for explicitly. Double tagging — the ID installed via both a plugin and the theme — inflates pageviews and halves your engagement metrics. Missing pages — the tag present on the homepage but not on a checkout or landing page built outside the CMS — silently breaks your funnel. Click through every important template once while watching Realtime.
⏱ 4 min read
Half of GA4 confusion is terminology. Five definitions clear most of it.
| Term | Definition |
|---|---|
| User | A unique visitor, identified by a cookie or signed-in ID. Devices and browsers split one human into several. |
| Session | A group of events from one user in one visit. Ends after 30 minutes of inactivity by default. |
| Event | Any recorded interaction: page_view, scroll, click, form_submit, purchase. |
| Key event | An event you flagged as a business outcome — GA4's replacement for the old Goals. |
| Engaged session | A session lasting over 10 seconds, OR containing a key event, OR with 2+ pageviews. |
Engagement rate is engaged sessions divided by total sessions. Bounce rate still exists, but it is now simply the inverse — the share of sessions that were not engaged — and neither metric appears in most reports until you add it via Customize report. That is deliberate: GA4 wants you looking at engagement time and key events instead.
The practical consequence is that a "high bounce rate" panic no longer applies the way it used to. A visitor who reads your article for four minutes and leaves is engaged, not bounced.
Automatically collected events fire with the base tag (page_view, session_start, first_visit). Enhanced measurement events come from that toggle (scroll, click, file_download, video_start, view_search_results). Recommended events are Google's standard names with standard parameters — use them when one fits, because reports understand them. Custom events are yours: name them in snake_case, action-first, like form_submit_contact.
Every GA4 report is dimensions crossed with metrics. A dimension is a description — page path, country, device, session source. A metric is a number — sessions, users, event count, key events. Rows are dimensions; columns are metrics. Once you see the pattern, building an exploration stops being mysterious: choose what you want to break down by, then what you want counted.
You can add a secondary dimension to most standard reports, which is where a lot of quick answers live. Landing page crossed with device category tells you whether your mobile experience is the problem. Traffic source crossed with landing page tells you which channel feeds which content.
GA4 also caps things. A property allows a limited number of custom dimensions and key events, and event names are case-sensitive — Form_Submit and form_submit are two different events that will split your reporting in half without warning. Neither limit bites a small site, but both reward deciding on a convention early.
Custom event names should be lowercase snake_case, action first: form_submit_contact, click_phone_header, download_pricing_pdf. Decide the convention before you create the second event, write it down, and apply it across every property you manage. Six months in, a property with thirty inconsistently named events is genuinely hard to report on, and renaming does not rewrite history.
⏱ 4 min read
GA4's Reports section is organised into collections — Life cycle (Acquisition, Engagement, Monetisation, Retention) and User (Demographics, Tech). You can ignore most of it. Four reports carry the weight for a typical site.
Reports › Acquisition › Traffic acquisition. Where sessions came from, grouped into channels: Organic Search, Direct, Referral, Organic Social, Paid Search, Email. This is the report to open when someone asks "is the SEO working?" — filter to Organic Search and compare periods.
Note the distinction alongside it: User acquisition credits the source that first brought a person to the site; Traffic acquisition credits the source of each session. For most questions you want traffic acquisition.
Reports › Engagement › Landing page. Which pages people arrive on, and how those sessions perform. This is the report that pairs with Search Console: GSC tells you what query brought them to the page, this tells you what happened next. The pages driving the most entries and the pages driving the most key events are frequently not the same pages, and that gap is where the useful work lives.
Reports › Engagement › Pages and screens. Every page by views and average engagement time. Add engagement rate through Customize report if you want it visible by default. Use it to find content that gets read versus content that gets loaded.
Thirty minutes of live activity. Genuinely useful for exactly two things: verifying that tracking works after a change, and watching a campaign or newsletter land. Not a reporting tool.
A 5-Minute Overview of GA4 Reports — Google Analytics. A compact tour of what each standard report is for. Open on YouTube ↗
Any standard report accepts comparisons — side-by-side views of subsets of your data. Add one for "Session source / medium exactly matches google / organic" and every report becomes an organic-only report without rebuilding anything. Two or three comparisons at once is the practical limit before a report becomes unreadable.
This is how you answer the question clients actually ask: not "how much traffic did we get?" but "is the organic traffic behaving differently from the paid traffic?"
Most standard reports hide metrics you want. Click Customize report (Editor or Administrator role required), add engagement rate or your key event as a column, and save it — either as a change to the existing report or as a new one in the left-hand navigation. Do this once per property for the Landing page report and you remove a repeated five-click detour from every future session.
Default to complete weeks or complete months. A Monday-to-Sunday comparison avoids the weekday-versus-weekend distortion that makes a seven-day window ending on a Tuesday look dramatic. For anything seasonal, compare year over year rather than month over month.
⏱ 4 min read
Without at least one key event, GA4 is a visitor counter. With them, it measures whether the site does its job.
A key event is just an ordinary event you flagged as important. Any event can be promoted — Admin › Events, toggle "Mark as key event". The name is new (they were called conversions until 2024); the concept is the old Goals, made simpler.
| Site type | Sensible key events |
|---|---|
| Service business / agency | Contact form submit, phone click, quote request, booking |
| Ecommerce | purchase, add_to_cart, begin_checkout |
| Content / publisher | Newsletter signup, account creation, 90% scroll on key pages |
| Lead gen | Form submit, PDF download, demo request |
Five to ten is the working range. Mark twenty things as key events and nothing is key. If an event does not change a decision you would make about the business, leave it as an ordinary event — it is still recorded and still available in explorations.
Three routes, easiest first. Some platforms emit useful events already — Shopify sends ecommerce events, many form plugins fire on submit. Failing that, GA4's Create event feature can build a new event from an existing one, for example turning a page_view of /thank-you/ into generate_lead. For anything more precise, use Google Tag Manager to fire an event on a real click or form submission.
The thank-you-page trick is the fastest reliable setup for most service sites, and it is worth building your forms to redirect to one for exactly that reason.
When a visitor arrives from organic search, leaves, then returns a week later from a newsletter and converts, which channel gets the credit? GA4's default is data-driven attribution, which distributes credit across touchpoints using your own historical data. The alternative is last-click, which gives everything to the final source.
Neither is wrong — they answer different questions — but it explains why GA4's conversion counts rarely match the numbers inside Google Ads or a social platform, each of which claims credit under its own rules. Pick one source of truth for reporting and note the discrepancy rather than trying to reconcile it.
Attribution only counts touchpoints inside a lookback window, configured under Admin › Attribution settings. Thirty days suits most transactional businesses. For B2B services with a long consideration period, ninety days is more honest — otherwise the organic article someone read six weeks before enquiring gets no credit at all, and you end up under-investing in exactly the content that works.
⏱ 4 min read
Standard reports answer standard questions. Explorations (the Explore section) answer yours. They are free-form: drag dimensions and metrics onto a canvas and build the table you wanted.
Explorations are limited by your data retention setting, which is the second reason to have changed it to 14 months in module two.
| Question | Tool |
|---|---|
| What did they search to find me? | Search Console |
| Is my page indexed? | Search Console |
| Which channel drives sessions? | GA4 |
| Which landing page produces leads? | GA4 |
| Why did organic traffic drop? | Both — GSC for cause, GA4 for consequence |
They will never match exactly. GSC counts clicks on a search result; GA4 counts sessions that loaded a tag. Ad blockers, bounces before the script loads, consent choices and session rules all open a gap. Compare directions, not absolute numbers.
GA4's interface is for analysis, not for clients. Looker Studio connects to GA4 and Search Console for free and lets you build a one-page dashboard with the four or five numbers that matter, refreshed automatically and emailed on a schedule. Build one template, duplicate it per client, and monthly reporting becomes a ten-minute review instead of an afternoon of screenshots.
Keep it short. A dashboard with forty charts gets ignored exactly as thoroughly as no dashboard at all.
⏱ ~6 min
Answer all ten, then score yourself. Explanations appear after scoring.
All consulted September 2026. Search and AI products change fast — check the primary documentation before acting on anything time-sensitive.