GA4 for Marketers: Setup, Key Events, Reports and Common Mistakes

Google Analytics 4, or GA4, is Google's analytics service that records what people do on your website and app as events, so you can see where visitors come from, what they do and which actions lead to enquiries or sales. To use it well, you need to define the actions that matter, mark them as key events, check that tracking works, and learn a small number of reports. Most marketers need only a handful.

GA4 replaced Universal Analytics and changed the model. Instead of sessions and page views as the core, everything is an event. That flexibility confuses many people, and a lot of GA4 properties are set up poorly, with missing conversions, noisy data and reports nobody reads. This guide explains the GA4 concepts that matter, a setup checklist, the reports a marketer should use, how to analyse acquisition and conversions, privacy and consent considerations, how to link GA4 with Search Console and ads, and the mistakes to avoid. We stick to documented behaviour from Google's help pages, which change regularly, so check them when in doubt. For the wider measurement picture, see our KPI dashboard guide and attribution models guide.

The GA4 model in plain words

Term	Meaning	Example
Event	Any interaction recorded as a data point	page_view, scroll, click, form_submit, purchase
Parameter	Extra detail attached to an event	Page title, product name, value, currency
Key event	An event you mark as particularly important to the business	generate_lead, purchase, booking_confirmed
User	A visitor or app user, identified by cookies or other identifiers	A person who may visit several times
Session	A group of interactions in a period of activity	One visit
Dimension	An attribute used to group data	Source, medium, country, device, page
Metric	A number measured	Users, sessions, engaged sessions, key events, revenue

Google's help page on key events explains that a key event is an event that measures an action that is particularly important to the success of your business. Any event you collect can become a key event: you identify or create the event for the action, then mark it as a key event. These replaced what GA4 earlier called conversions. Source: Analytics Help, About key events.

Step 1: Decide what to measure before you touch settings

1. List your business goals. Online sales, enquiries, bookings, calls, downloads.
2. Define each as a specific action a visitor can take, such as submitting the enquiry form or reaching the booking confirmation page.
3. Rank them. Mark the two or three that most reflect revenue as your main key events. Keep lower-value actions as ordinary events.
4. Write a measurement plan. Event name, trigger, parameters and owner for each.

The most common GA4 mistake is marking too many events as key events, so the report shows lots of "conversions" that do not mean anything. Keep the list short.

Step 2: Set up the property properly

• Create a GA4 property and a web data stream for your site and install the tag, usually through Google Tag Manager or your CMS.
• Turn on enhanced measurement for automatic events such as scrolls, outbound clicks and file downloads, and check which events you actually want.
• Send your key events. For purchases, use the recommended ecommerce events with values and currency. For leads, send a clear lead event when the form is successfully submitted, not when the button is clicked.
• Exclude internal traffic from your offices and your agency. Create filters in the admin settings.
• Add hostname filters. Google added Include data filters for hostnames in September 2026, so you can allowlist domains authorised to send data and block spam.
• Set data retention to a period that suits your needs and policies. Google notes that Google signals data is retained for a maximum of 26 months regardless of your settings.
• Set your reporting identity, time zone and currency.
• Control access. Give people the lowest role they need.
• Connect Search Console and, if used, Google Ads. See below.

Step 3: Handle consent and privacy

Analytics uses cookies and identifiers, so privacy law applies. In the UK, the rules on storing or accessing information on a user's device generally require consent for non-essential cookies, and UK GDPR applies to personal data. India's DPDP Rules require clear consent notices where consent is the basis. Use a consent management tool, configure Google's consent settings so tags behave according to the choices people make and describe your analytics use in your privacy notice. Be aware that consent choices reduce the data you can see, so your numbers will undercount real behaviour. This is general information, not legal advice. Take advice for your situation.

Step 4: Test that tracking works

1. Use the realtime report and the debug view to confirm events fire when you perform actions.
2. Complete a real test conversion, such as submitting a test enquiry or a test purchase, and check it appears once, with the right values.
3. Check that events fire on each page type and each device.
4. Check the same conversion is not counted twice on page refresh.
5. Re-test after every website change, as updates often break tags.

Many weak marketing decisions rest on broken tracking. Testing is cheap.

The reports a marketer actually needs

Question	Where to look	What to read
Where do visitors come from?	Reports, Acquisition: Traffic acquisition	Sessions, engaged sessions and key events by channel and by source and medium
Which pages bring people in and which convert?	Reports, Engagement: Landing page	Sessions, engagement rate, key events and revenue by landing page
What do people do on the site?	Reports, Engagement: Pages and screens, Events	Views, engagement time and event counts
Which actions lead to goals?	Reports, Engagement: Key events	Counts and trends of your key events
How do sales perform?	Reports, Monetization	Revenue, transactions and products, if ecommerce events are set up
How does credit across channels look?	Advertising: Attribution	Attribution models and paths. See below
Where do people drop off in a journey?	Explore: Funnel exploration	Step-by-step completion
Custom views for your team	Dashboards, introduced in September 2026	Your own selection of KPIs

Use the Explore section for custom analysis, such as funnels and path analysis, but do not start there. Master the standard reports first.

Understanding engagement

GA4 reports engaged sessions and engagement rate, which replaced bounce rate as the main indicator of whether a visit did anything meaningful. An engaged session is one that lasts longer than a set time threshold, includes a key event or includes more than one page or screen view, according to Google's definitions, and the time threshold can be adjusted in settings. Use engagement alongside key events. High engagement with no conversions may signal a page that informs but does not persuade, while low engagement from a source may signal mismatched traffic.

Analysing acquisition

Channel groups

GA4 groups traffic into default channels such as organic search, paid search, direct, email, referral and social. Misclassified traffic is a frequent problem, for example unlabelled email links showing as direct. Use UTM parameters on links you control, such as email and social posts, so GA4 can attribute them correctly.

UTM parameters

Parameter	Use	Example
utm_source	Where the link was placed	newsletter
utm_medium	The type of channel	email
utm_campaign	The specific campaign	autumn_offer
utm_content	Differentiates links in the same campaign	top_button

Use consistent lowercase names and a shared naming sheet. Never use UTM parameters on internal links, because they reset the source of the session.

AI referral traffic

Visits from AI products appear as referrals. OpenAI says ChatGPT referral links include utm_source=chatgpt.com, so you can filter session source for chatgpt.com, and other products such as Perplexity usually show as referrals from their own domains. See our guides to ChatGPT search and Perplexity. Google's AI Overviews and AI Mode clicks are included in overall Google search traffic, and the Generative AI performance report in Search Console shows impressions.

Linking GA4 with other tools

• Search Console. Link it to see queries and landing page performance alongside engagement and conversions. Be aware the two tools count differently and will not match exactly.
• Google Ads. Link accounts to import key events as conversions and build audiences, and make sure conversions you import are the ones you want to bid on.
• BigQuery. Export raw event data for advanced analysis if you need it.
• Your CRM. Pass lead IDs or import offline results so you can judge quality, not just quantity.

Attribution in GA4

Google's help page on attribution explains that GA4 offers three attribution models in its attribution reports: data-driven attribution, paid and organic last click and Google paid channels last click. It notes that all models exclude direct visits from receiving credit unless the path to the key event consists entirely of direct visits. Data-driven attribution uses machine learning on your account's data to distribute credit, evaluating both converting and non-converting paths. Google has also retired first-click, linear, time decay and position-based models from its reports. We explain how to use attribution sensibly, and its limits, in our guide to marketing attribution models. Source: Analytics Help, Get started with attribution.

A simple GA4 routine for small businesses

When	Task
Weekly (10 minutes)	Check traffic acquisition and key events for sudden changes. Check realtime after any site change
Monthly (1 hour)	Review landing page performance and conversions by channel. Note what changed and why. Share one page of findings
Quarterly	Review the measurement plan, key events, filters and consent. Remove unused events
After every website change	Test tracking again

Common GA4 mistakes

• Marking every event as a key event.
• Not testing conversions after setup or site changes.
• Counting a button click as a lead, instead of a successful submission.
• Duplicate tags, which double-count events.
• No internal traffic filter, so staff visits skew data.
• Untagged campaign links that fall into direct.
• Comparing GA4 with other tools and expecting identical numbers.
• Ignoring consent, so data is incomplete or unlawful.
• Building complex explorations before understanding standard reports.
• Reporting traffic without business results.

A worked example: fixing a broken lead tracking setup

This is an invented example. A Cardiff accountancy firm reports 240 "conversions" a month in GA4, but its partner says only about 25 real enquiries arrive. Here is how the firm fixes it.

Diagnose

• In the key events report, three events are marked as key events: click (for any outbound link), scroll and form_submit. Together they produce the 240.
• The form_submit event fires when the button is clicked, even if the form shows an error and the enquiry is not sent.
• Internal visits from the office are not excluded.

Fix

1. Define the goal. A "qualified enquiry" is a successfully submitted contact or quote form, or a click on the phone number from a mobile device.
2. Change the form event. A developer sends generate_lead only when the confirmation message appears, with a parameter for the form name.
3. Track calls. A click on the phone link sends a phone_click event, and the firm uses a call tracking number to count real calls.
4. Update key events. Only generate_lead and phone_click are marked as key events. click and scroll are left as ordinary events.
5. Exclude internal traffic using the office IP address.
6. Test. Staff submit a test enquiry from a phone and a laptop, and confirm one event each in the realtime and debug views.
7. Record the change date, and note that data before it is not comparable.

Result

The next month shows 31 key events, which is close to the partner's count of 25 real enquiries plus a few duplicates and test records. The numbers are smaller but trustworthy, and decisions improve because the team can finally see which landing pages and sources produce real enquiries.

An event naming convention

Consistent names make reports easier. A simple convention for custom events:

Rule	Example
Use lowercase with underscores	generate_lead, phone_click
Use recommended event names where one exists	purchase, add_to_cart, generate_lead
Describe the action, not the page	quote_request instead of contact_page_button
Add parameters for detail	form_name, service_type, value
Document every event	A shared sheet with name, trigger, parameters and owner
Avoid personal data in parameters	No names, emails or phone numbers in event data

Do not send personally identifiable information to analytics. Google's terms prohibit it, and privacy law applies.

Reading the numbers: a quick guide

Question	Look at	Be careful about
Did traffic rise?	Sessions and engaged sessions by channel over a comparable period	Compare like with like, including seasonality and consent changes
Is traffic good quality?	Engagement rate and key events per session by source	A source with few sessions can look good or bad by chance
Which pages convert?	Landing page report: key events and revenue per session	Pages that attract different traffic types are not directly comparable
Did the campaign work?	Sessions and key events with the campaign UTM	Untagged links and delayed conversions undercount
Where do people drop off?	Funnel exploration	Define each step carefully and test the events

Limits and surprises to know about

• Thresholds. Some reports may hide data to protect user privacy, for example when audience data could identify individuals.
• Consent and modelling. Data from users who decline cookies may be missing or modelled, depending on your settings.
• Processing delay. Some data takes hours to appear in standard reports.
• Row and cardinality limits. Reports with very many unique values can group the rest into "other".
• Sampling in explorations. Large queries may use sampled data.
• Differences with other tools. Ad platforms and Search Console count differently. Do not expect matches.

Common GA4 questions

Question	Answer
What replaced bounce rate?	Engagement metrics: engaged sessions and engagement rate. GA4 treats a session as engaged if it lasts longer than a set time, includes a key event or includes more than one page or screen view
Why are there fewer conversions than before?	You may have fewer events marked as key events, consent may reduce data, or tracking may have been fixed. Check your measurement plan
How do I track phone calls?	Track taps on phone links as events, and use call tracking numbers or imports if you need to know which calls became customers
Can I see individual users?	GA4 is designed for aggregated analysis. Do not try to identify individuals, and respect privacy law
Do I need Tag Manager?	Not required, but it makes tags and events easier to manage without code releases
How do I compare periods?	Use the date comparison control, and compare like with like, accounting for seasonality

A launch and audit checklist

• Data stream and tag installed once, without duplicates.
• Enhanced measurement settings reviewed.
• Key events limited to meaningful actions, with values where relevant.
• Internal traffic filter active and hostname filters considered.
• Data retention and reporting identity set deliberately.
• Consent mode and banner tested, with the effect on data understood.
• Search Console and Google Ads linked where used.
• UTM naming convention documented.
• Roles and access reviewed, with least privilege.
• A test conversion completed and verified end to end.
• A documented measurement plan and change log.

Using GA4 with a CRM

For lead generation, GA4 shows how many form submissions occurred. Your CRM shows which leads became customers. Connect them:

1. Capture the source, medium and campaign on each lead with hidden form fields.
2. Store them in the CRM record.
3. Report on qualified leads, opportunities and revenue by source.
4. Where platforms allow, import offline conversions, so ad systems learn from real outcomes.
5. Respect consent and privacy rules throughout.

A story of a GA4 report that led to a better page

This is an invented composite. A Liverpool company selling commercial kitchen equipment saw steady traffic to its category pages but few enquiries. The marketing manager opened the landing page report in GA4, set the date range to the last quarter and added the key event column for "quote requests". Some pages had many sessions but few quote requests, and one stood out: the page for commercial dishwashers had the second-highest traffic and almost the lowest rate of quote requests.

She segmented the report by device. On desktop, the quote rate was reasonable. On mobile, it was less than half. She then used the funnel exploration to see the path from landing page to quote form submission and found that many mobile visitors reached the form page but did not complete it. She checked the page speed report and saw that the form page loaded slowly on mobile, partly because of a large banner image and a third-party chat widget.

She also used the site search report, which showed that visitors on the dishwasher page often searched for "dimensions", "water usage" and "installation requirements", which were missing from the page. The page showed product photos and a long marketing paragraph but no specifications table.

The team made changes. It compressed the banner image, delayed the chat widget until after the page loaded, shortened the form to five fields and added a clear specifications table with dimensions, power and water requirements, noise levels and installation notes, together with a "download the specification sheet" link. It checked that the form event fired only after a successful submission, and tested it on three phones.

Over the next two months, the mobile quote rate for the page rose substantially, and total quote requests increased. The manager recorded the date of the change in a log, and also noted that she could not isolate the effect of each change. But the process showed what GA4 does best when used well: it helps you find where the problem is, in this case a specific page and device, and points toward questions to investigate with other evidence, such as site search and page speed. It did not give the answer by itself. The answer came from combining data with common sense about what a buyer needs to decide, and from testing the fix on real devices.

Where we can help

We are a digital marketing agency in Manchester, UK and Mumbai, India. We audit GA4 setups, fix tracking and build simple dashboards tied to revenue, as part of our CRO and PPC management work. Contact us for a tracking review.