Updated 2nd September 2026

A practical guide to Google Analytics 4: Get the most out of GA4

Google Analytics 4 (GA4) is where most businesses find out what's actually working - which channels bring visitors, what those visitors do, and where sales, leads and sign-ups really come from. Used well, it's the difference between marketing decisions based on evidence and decisions based on gut feel.

This practical guide walks through the essentials - the event model, the metrics that matter, customising reports, reading your traffic sources - then goes deeper on three areas that repay the effort: the Advertising section, audiences, and e-commerce tracking.

How GA4 works: The event model

Google retired Universal Analytics in July 2023 and replaced its session-and-pageview model with something more flexible. In GA4, everything a visitor does is an event. A page load is an event. So is a scroll, a video play, a click on an outbound link, an added-to-basket product and a completed purchase. Each event carries parameters - the details that describe it, like the page title, the product ID or the transaction value.

Strengths and weaknesses

Where GA4 is strong

• Free and generous - the standard tier costs nothing and includes the BigQuery export that was previously reserved for paying enterprise customers.
• Measures what matters - the event model captures any interaction you care about, from a demo request to a basket abandonment.
• Cross-device journeys - where signed-in data allows, GA4 can follow the same user from a phone browse to a desktop purchase.
• Built for a privacy-first web - consent mode support and modelling designed for a world with fewer third-party cookies.
• Machine learning signals - predictive metrics like purchase probability and predicted revenue can power smarter remarketing.
• Deep custom analysis - the Explore workspace goes far beyond anything the standard reports offer.

Where it frustrates

• The learning curve is real - new terminology, new navigation and a very different logic to Universal Analytics.
• Sparse defaults - the out-of-the-box reports are thinner than their UA equivalents, so expect to customise.
• Your UA history is gone - Google deleted Universal Analytics data in July 2024. If it wasn't exported, it can't be recovered.
• Thresholding and sampling - GA4 sometimes hides or samples data on smaller segments to protect privacy, which can frustrate granular analysis.
• Gaps get papered over - when consent is declined or attribution breaks down, GA4 fills the blanks with modelled estimates and pushes unattributed visits into 'Direct'.
• It won't match Google Ads exactly - different counting methods and attribution mean conversion numbers differ between the two platforms.

Events and key events explained

Since everything in GA4 is an event, the platform makes more sense once you understand the three layers of events you'll work with.

Automatically collected and Enhanced measurement events - GA4 tracks page_view, session_start, first_visit, scrolls, outbound clicks, site search and video engagement out of the box (toggle these under your web data stream's Enhanced measurement settings).

Custom events - the interactions specific to your business, usually deployed through Google Tag Manager: a quote request, a brochure download, a click-to-call.

Key events - GA4's name for what used to be called conversions. These are the events with commercial value: purchases, leads, sign-ups, bookings. Flag them in Admin > Events with the 'Mark as key event' toggle. Once flagged, they appear throughout the Advertising section and, if your accounts are linked, become available in Google Ads as conversion actions your campaigns can optimise towards.

Users, sessions and engagement: What the numbers mean

GA4's headline metrics look interchangeable but measure different things. Getting these definitions straight avoids a lot of confused reporting.

• Total users - everyone who triggered any event in the date range.
• Active users - the default 'Users' figure in most reports. Only counts users with an engaged session, meaningful activity or a key event, filtering out the bounce-in-two-seconds noise.
• Sessions - a burst of activity from one user, ending after 30 minutes of inactivity. One user can rack up many sessions.
• Engaged sessions - sessions lasting 10+ seconds, containing a key event, or covering at least two pages.
• Engagement rate - engaged sessions as a share of all sessions. Roughly the mirror image of UA's old bounce rate.
• Average engagement time - time the page was actually in focus, not just open in a background tab.

Engagement metrics are a better quality signal than raw pageviews. A landing page with high views and low engagement time is telling you something - usually that the traffic is wrong or the page isn't delivering on its promise.

Finding your way around

The left-hand rail organises GA4 into five working areas:

• Home - a configurable dashboard of headlines and automated insights.
• Reports - the standard reports: Acquisition, Engagement, Monetisation and Retention under Life cycle, Demographics and Tech under User, plus Realtime.
• Explore - the custom analysis workspace.
• Advertising - attribution and conversion journey reporting (more on this below).
• Admin - data streams, events, key events, audiences, custom dimensions, linked accounts and data settings.

Customising reports and adding breakdowns

Editing a standard report

Open any standard report and click the pencil icon (top right) to enter Customize mode. You can add or remove dimensions and metrics, change charts, reorder the summary cards and apply filters. Save over the original or use 'Save as a new report' to keep both. You'll need Editor permission on the property.

The secondary dimension trick

The fastest way to get more from any report is a breakdown column:

1. Open a report - 'Pages and screens' under Engagement is a good example.
2. Click the blue '+' next to the primary dimension column header.
3. Choose a secondary dimension - 'Session source / medium', 'Device category' and 'Country' are the workhorses.

Now one table answers questions like 'which channels drive traffic to our best-converting landing pages?'. Remove the breakdown with the 'x' on the column.

Reading traffic sources: Source, medium and channel

GA4 labels every visit with a source (where it came from - 'google', 'linkedin.com', 'newsletter') and a medium (what kind of traffic - 'organic', 'cpc', 'email', 'referral'). Combinations roll up into 

default channel groups - Organic Search, Paid Search, Organic Social, Paid Social, Email, Direct, Referral and so on - for tidier reporting.

Typical combinations you'll see:

• 'google / organic' - an unpaid Google search click.
• 'google / cpc' - a Google Ads click, auto-tagged through the account link.
• '(direct) / (none)' - a typed URL, a bookmark, or any visit GA4 couldn't attribute.
• 'linkedin.com / referral' - an untagged LinkedIn click, attributed by referring domain.
• 'newsletter / email' - an email link tagged with UTM parameters.

Tag every paid, email and social link with UTM parameters. Untagged links leak into 'Direct' and 'Referral' and quietly undercount your best channels.

Which source dimension should you use?

GA4 scopes source / medium three ways, and the scope changes the answer. First user source / medium records where someone came from on their very first visit - use it for acquisition analysis. Session source / medium records where each individual session came from - use it for day-to-day campaign reporting. Event-scoped source / medium attributes individual events. When in doubt, Session source / medium is the one you want.

The advertising section: Attribution and user journeys

The Advertising section (in the left rail) is the most underused part of GA4. It's where you stop asking 'which channel got the last click?' and start understanding how channels work together to produce a conversion. Everything here is driven by your key events, so it only becomes useful once those are set up properly.

Performance and all channels

The Performance report breaks conversions and revenue down by channel, source / medium and campaign for your paid activity. The All channels view widens this to every channel, paid and organic - useful for seeing how email, organic search and direct traffic contribute alongside your ad spend, all measured against the same key events.

Attribution models and model comparison

An attribution model is the rule deciding which touchpoint gets credit for a conversion. GA4's default is data-driven attribution, which uses machine learning to share credit across the touchpoints that actually influenced the outcome, rather than handing everything to the last click.

The Model comparison report lets you view the same conversions through different lenses side by side - data-driven vs 'paid and organic last click' vs 'paid and organic first click'. If a channel's conversions jump when you switch from last click to first click, that channel is doing introduction work your last-click reports have been hiding.

Conversion paths

Conversion paths shows the actual sequences of touchpoints that led to each conversion, split into early, mid and late positions. You'll see patterns like 'Paid Social > Organic Search > Direct > purchase' with counts and revenue attached, plus how much credit each channel earned under your chosen model.

Read it as a team sheet rather than a leaderboard. Some channels are openers that introduce your brand, some are assists that nudge people along, and some are closers that finish the job. Cutting a channel because it rarely closes can quietly break journeys further up the funnel - Conversion paths is where you check before you cut.

Attribution settings worth checking

In Admin > Attribution settings you can change the reporting attribution model and the conversion windows (how long after a touchpoint a conversion can still be credited - up to 90 days for acquisition events). Set these deliberately, and note them in your reporting, because they change your numbers.

Building and using audiences

Audiences are saved groups of users who match conditions you define. They're built in Admin > Audiences, either from Google's suggested templates or from scratch with the condition builder, and they're one of the most commercially useful features in GA4.

What you can build

• Visitors who viewed specific products or pages without buying
• Basket abandoners - added to cart, never reached purchase
• High-intent researchers - multiple sessions and product views in a short window
• Existing customers, for exclusion from acquisition campaigns or targeting with upsell offers
• Lapsed customers who haven't returned within a set number of days
• Predictive audiences - Google's machine learning builds groups like 'likely 7-day purchasers' and 'likely 7-day churning users' once your property has enough conversion data.

Sequences, exclusions and membership

The condition builder goes deeper than most people realise. Sequence conditions capture users who did things in a specific order ('viewed a product page, then returned within 7 days'). Exclusion conditions strip out users you don't want ('has not purchased'). Membership duration controls how long a user stays in the audience after qualifying - up to 540 days, though shorter windows usually make for sharper remarketing.

Putting audiences to work

An audience does three jobs. In GA4 itself, use audiences as comparisons in any report - overlay 'basket abandoners' on your traffic reports and see which channels create them. In Google Ads, linked audiences flow through automatically for remarketing and as signals that sharpen Smart Bidding and Performance Max targeting. And audience triggers can fire an event when someone joins an audience, letting you count 'became a high-intent visitor' as a key event in its own right.

A note on scale: audiences need at least 100 members before they can be served ads, and they populate from the moment they're created (they don't backfill historic users). Build your core audiences early, even if you won't use them for a month.

E-commerce tracking in depth

GA4's Monetisation reports are where online retailers live, but they only work as well as the event data feeding them. Getting the full event set implemented - not just the purchase - is what unlocks funnel analysis, merchandising insight and accurate revenue attribution.

The events that matter

Google defines a recommended set of e-commerce events. The core sequence is view_item_list (a category or search results page), select_item (clicking a product), view_item (the product page), add_to_cart, view_cart, begin_checkout, add_shipping_info, add_payment_info and finally purchase. Each carries an items array describing the products involved - item_id, item_name, price, quantity, item_category, item_brand and optionally item_variant, discount and coupon.

Most platforms handle this for you: Shopify, WooCommerce, Magento and BigCommerce all have native integrations or well-supported plugins, typically wired through Google Tag Manager and a dataLayer. Whatever the setup, verify it in DebugView (Admin > DebugView).

The reports you unlock

• E-commerce purchases - revenue, units and average price by item, category and brand. This is your merchandising view: what actually sells, at what price, from which lists.
• Purchase journey - the standard funnel from session start through view_item, add_to_cart and begin_checkout to purchase, with drop-off rates at every step. Segment it by device or channel and weak steps stand out immediately.
• Checkout journey - the same idea inside checkout itself: where buyers stall between shipping, payment and confirmation.
• Promotions and item lists - if you send promotion and item_list parameters, GA4 reports which on-site placements and campaigns actually drive views and sales.

Making revenue data work harder

Purchase revenue flows through everything once it's tracked: the Pages and screens report gains a revenue column, Advertising reports attribute revenue (not just conversion counts) across channels, audiences can be built on spend thresholds, and linked Google Ads campaigns can optimise to revenue through Smart Bidding rather than treating a £15 order and a £1,500 order as equal wins.

Explore: Custom analysis

When the standard reports run out of road, Explore is the workspace for building your own. Drag dimensions, metrics, segments and filters onto a canvas and assemble exactly the view you need.

• Free form - pivot-table style analysis over any of your data.
• Funnel exploration - define any sequence of steps and measure drop-off, segmented however you like.
• Path exploration - follow the actual routes users take, forwards from a landing page or backwards from a purchase.
• Segment overlap - see how audiences intersect, e.g. mobile users vs paid traffic vs purchasers.
• User explorer - inspect individual anonymised journeys when aggregate data isn't enough.

Explorations can be saved and shared with your team, and they're where most serious GA4 analysis ends up.

Want help setting GA4 up properly, or getting more out of the data you already have? Get in touch with the team at Atelier Studios.

FAQs - Google Analytics 4

What replaced conversions in GA4?

Key events. Any tracked event can be flagged as a key event in Admin > Events, and flagged events flow into the Advertising reports and into linked Google Ads accounts as conversion actions.

Which attribution model should I use?

Data-driven attribution (the default) is right for most businesses - it distributes credit based on observed influence rather than position. Use the Model comparison report to understand how it differs from last click before you rely on it in reporting.

Why is so much of my traffic 'Direct'?

Direct is GA4's bucket for visits it can't attribute - typed URLs and bookmarks, but also untagged email and social links, some app traffic and visits where consent or privacy features stripped the referrer.

Do I need a developer to set up e-commerce tracking?

On the major platforms, usually not - Shopify, WooCommerce and similar handle the recommended events through integrations or plugins. Custom-built sites need a developer to populate the dataLayer. Either way, verify the events and values in DebugView before trusting the numbers.

How many users does an audience need before I can use it?

Audiences need at least 100 active members before ads can serve to them. They also only collect members from the moment of creation, so build your key audiences early.

Why don't GA4 and Google Ads report the same conversions?

Google Ads counts at the time of the ad click using its own attribution and windows, GA4 counts at the time of the conversion with its own model. Cross-device journeys, time zones and consent modelling widen the gap. Pick one platform as your source of truth per metric and be consistent.

How long does GA4 keep my data?

Event-level data is kept for 2 months by default, extendable to 14 months in Admin > Data Settings > Data Retention - worth changing on day one. Aggregated data in standard reports is retained indefinitely.