How to Configure GA4: Events, Key Events and Rollout

A clean GA4 setup rests on four building blocks: a tracking plan with consistently named events, a handful of genuine key events, consent handling that takes effect before the first tag fires, and a tested, documented rollout. Follow that order and you get numbers that management and sales can rely on; just paste in the Google tag and you get data chaos.

Since Universal Analytics was shut down in 2023, Google Analytics 4 (GA4) has been Google’s analytics tool for websites and apps. It no longer has hit types such as pageview or transaction: every interaction is an event with parameters. That is flexible but demands planning – especially in B2B, where an inquiry often arrives weeks after the first visit and involves several people.

The GA4 data model: properties, data streams and events

A property bundles the data of one offering, and data streams feed it from your website or app. Decide up front whether website and app share a property, and keep test and live data apart, for example with a separate test property. Internal traffic from the office, home offices or service providers should be excluded from day one with a data filter; you can check the filter in testing mode before it becomes active.

GA4 distinguishes four kinds of events:

• Automatically collected events: such as first_visit or session_start
• Enhanced measurement: scrolls, outbound clicks, file downloads, video engagement, site search and form interactions can be switched on with a toggle
• Recommended events: names and parameters predefined by Google, for example generate_lead, sign_up, purchase or begin_checkout
• Custom events: your own names for actions Google does not provide for, such as a click on a callback button

Enhanced measurement is a good start, but it is not a plan. The form event, for instance, can fire as soon as a form is submitted – even if validation fails or the data never arrives. That makes it unsuitable as proof of an inquiry. A dedicated event that is only sent after a successful submission is more reliable, for example via the website’s data layer. Switch off automatic events that overlap with your own.

Planning events: naming convention and tracking plan

Use the recommended events wherever possible – for B2B contact forms that is generate_lead – and add your own according to a fixed pattern: lowercase, underscores, verb first, no special characters. Event names are case-sensitive: “form_submit”, “formSubmit” and “contact_sent” are three different events for the same action. On top of that, standard properties have limits for event names, parameters and custom dimensions, and uncontrolled growth will eventually hit them.

The core of it is the tracking plan as a table: which event fires when, which parameters belong to it, and who is responsible? An example for a machine builder with a contact form, appointment booking, catalog download and callback button:

Event	Trigger	Key parameters	Key event
generate_lead	Contact or callback form submitted successfully	form_type, page_section	Yes
meeting_booked	Appointment booking confirmed	meeting_type	Yes
click_phone	Click on the phone number	link_location	Only if calls count as inquiries
file_download	Catalog PDF downloaded (enhanced measurement)	file_name	No
newsletter_signup	Sign-up form submitted	form_location	No

For custom parameters to appear in reports and explorations, register them in GA4 as custom dimensions. Just as importantly, record what must not be sent: names, email addresses and other personal data belong neither in event parameters nor in URLs – Google’s terms of use expressly prohibit this.

Key events instead of conversions: what really counts

In the GA4 interface, conversions have been called “key events” since 2024; the term “conversion” mostly remains in Google Ads for campaign management. Whatever the name, mark only genuine business goals – in practice usually four to eight per property – such as a contact inquiry, appointment booking, quote request or purchase. Scroll depth, video starts or downloads stay ordinary events. They help with analysis without diluting your target figures. Only then does the conversion rate become meaningful.

• Choose the counting method deliberately: a key event can be counted per event or once per session. Count purchases per event; for contact inquiries once per session is often enough so that repeated submissions do not inflate the number.
• Link with Google Ads: key events can be imported into Google Ads and used as an optimization goal. First check that the event genuinely means an inquiry and not just a click.
• Tag campaign links consistently: with UTM parameters for source, medium and campaign you attribute newsletters, ads and social posts cleanly. Without a uniform spelling, traffic ends up in “Direct” or “Unassigned”.

In B2B, decisions run across several people, devices and weeks; no analytics tool captures that completely. GA4 figures also rarely match ad platforms or the CRM exactly: attribution models differ, and data is missing without consent. So judge trends rather than single values, reconcile samples with the CRM – for example ten to twenty opportunities per quarter – and add an optional field “How did you hear about us?” to your contact form. It provides hints that no tracking captures. Which other metrics matter is covered in the article on web analytics and KPIs.

Privacy and consent from the start

GA4 sets cookies and accesses information on the user’s device. Section 25 of Germany’s TDDDG (the telecommunications and digital services data protection act) generally requires prior consent for this, and the processing of the data itself is also subject to the GDPR. In practice, this means the Google tag may only send measurement data after consent – or only to the extent your legal assessment allows for Google Consent Mode v2.

Consent Mode v2 passes the consent signals (analytics_storage, ad_storage, ad_user_data, ad_personalization) on to the Google tags. In basic mode, tags load only after consent; in advanced mode, signals without cookies are also sent when consent is missing. The latter is assessed differently under data protection law, so have the variant approved before you use it. These points are part of the basics:

• consent management platform with equally prominent buttons for “Accept all” and “Reject all”
• privacy policy that describes GA4 and the data recipients correctly, plus the data processing terms with Google
• set data retention deliberately – standard properties allow a maximum of 14 months for user-level data
• turn on Google signals and ads personalization only if you really need them
• test consent: reject, accept, withdraw – and check what arrives in DebugView each time

How tracking and trust can go together is covered in the article on GDPR-compliant tracking. Expect gaps: people who decline are not measured. That is exactly why trends and CRM reconciliation matter more than the decimal place.

GA4 rollout in seven steps

1. Clarify goals: which decisions should the data support, and which business goals count?
2. Take stock: for an existing property, check which events are running, which are used and which can go.
3. Write the tracking plan: record events, parameters, key events and owners in a table.
4. Implement data layer and tags: the website exposes relevant information through a data layer; tags run via Google Tag Manager or directly in code and are tied to consent.
5. Test: use DebugView and Tag Manager preview to check that all events arrive completely – with and without consent, on desktop and smartphone.
6. Document and go live: version every tracking change, keep the plan up to date and monitor closely during the first days after launch.
7. Analyze and maintain: align reports and explorations with the defined goals and review once a quarter which events are still needed.

Suppose a machine builder finds that GA4 reports far more contact inquiries than sales actually receives. The cause is the enhanced-measurement form event, which also fires on failed submissions, and the tag is loaded both in the code and in Tag Manager. After cleanup, a generate_lead event sent after a successful submission replaces both, and the figure is back in the same range as the CRM inquiries.

Common mistakes in GA4 setups

• Double counting: the tag is loaded in the code and additionally in Tag Manager, so page views are counted twice.
• Too many key events: if every click counts as a goal, the numbers lose their meaning.
• Broken sessions: with several domains or external booking and payment providers, sessions break unless cross-domain measurement and unwanted referrals are configured.
• Unregistered parameters: custom parameters are sent but appear in no report.
• Personal data in parameters or URLs: for example email addresses from forms or newsletter links.
• Missing documentation: nobody remembers why an event exists or who uses it.

Reporting and BigQuery without data chaos

Reporting needs only a few clearly defined views: one for management with the key goals, one for marketing with channels and campaigns and, if needed, one for sales showing the quality of inquiries. GA4 explorations suit deeper questions. If you want to combine raw data with other sources such as the CRM, you can link GA4 to BigQuery. That only pays off once the event structure is clean – a data warehouse does not fix a chaotic landscape, it only makes it bigger. Also note that exported data can still be personal data and needs a deletion concept.

How Nano Concepts helps

Good tracking needs technical understanding of data layers, Tag Manager and interfaces as well as marketing know-how for goals, campaigns and analysis. Nano Concepts handles the technical implementation as part of our web development; we carry out analysis and campaign management together with our partner agency LOMA DESIGN. If you want to review your existing setup or build a new one, we will clarify in a free initial consultation where you stand and what makes sense.

This article does not replace legal advice.