Traffic is growing in Google Analytics, the advertising system reports conversions, yet the business still sees no more qualified enquiries. Another report will not help at this point. First we need to find out what the individual numbers mean, where they came from and whether they support a decision.
I use GA4 as an analytics layer over a website and application, not as an accounting system or universal truth about a customer. It can connect events, traffic sources and selected business steps. Its value rests on the measurement plan and the quality of the implementation.
Start with a decision, not a list of metrics
Every report should answer a question that can change what you do next. For example:
- Which pages bring relevant enquiries?
- Where do people abandon an order?
- Which channel brings the first visit and which helps completion?
- Which types of content lead to work with the service?
- Which campaign consumes budget without a business signal?
Then choose events and parameters. Not the other way around. When you measure fifty clicks without a connection to a decision, you are only producing more expensive noise.
GA4 is built on events and their parameters
Page views, searches, enquiry submissions and purchases are events in GA4. Parameters add context: value, currency, transaction ID, form name or items.
Google distinguishes automatically collected, enhanced, recommended and custom events. When a recommended name and parameters exist for an action, I use them. For example, generate_lead for a submitted request for information and purchase for a completed purchase. This supports prepared metrics and future integrations.
Create a custom event only when the standard meaning does not fit. The name button_click_7 measures something, but in six months nobody will know what.
A key event must matter to the business
GA4 now calls important actions key events. Do not label every ordinary interaction this way. Scrolling, opening a menu and starting a video can be diagnostic events. An enquiry, registration or purchase can be a key result.
For a lead-generation website I distinguish:
- a successfully submitted form,
- a click on an email address or telephone number,
- a qualified contact confirmed in the CRM.
The first is technically measurable, the second is strong intent and the third is business quality. If we combine them into one number, marketing looks better but the company makes worse decisions.
E-commerce needs a stable transaction, not a thank-you page
For a purchase, use the recommended purchase event and its corresponding parameters, especially a stable transaction_id, value, currency and items according to your model. A unique ID helps limit duplicate purchases and connect analytics with the ordering system.
The event should be created when the business status is confirmed. Merely opening or refreshing a thank-you URL is not a reliable source. Test successful and failed payments, repeated loads, discounts, shipping, tax and multiple currencies. Finally compare the data with the backend. The difference need not be zero, but it must be explainable.
Enable enhanced measurement after checking it
Without custom implementation, GA4 can collect such things as scrolls, outbound clicks, site searches, interaction with embedded video or file downloads. It is a useful start. It does not mean that an automatic event matches your business definition.
Check that search recognises the parameter used by your website, that video uses a supported embedding method and that an automatic event does not duplicate your own GTM measurement. Turn off what you do not use.
Google Tag Manager helps only with stable data
I use GTM for events the website passes in the data layer and to manage multiple tags and conditions. I do not recommend extracting business values from page text or triggering a purchase from part of a URL when the application can send a real event.
How to design the data layer, consent, testing and versions is covered in the separate article Google Tag Manager: proper setup.
Consent changes which data you can observe
GA4 and advertising tags must respect the user’s choice. Google Consent Mode can pass consent status and adjust the behaviour of supported tags. It does not replace a cookie banner or legal assessment.
Google distinguishes a basic mode, in which tags are blocked before consent, and an advanced mode, in which supported Google tags load with a default denied state and may send limited cookieless signals. The choice has technical, legal and data consequences. I would not make it based on which option creates the nicer graph.
After deployment, test both consent and rejection. Check not only Google tags but also Meta, heatmaps, chat and every other service in the container.
DebugView confirms transmission, not business meaning
Google recommends checking events in Realtime and DebugView. DebugView shows events and parameters from a test device in the order in which they arrived.
During a check I watch for:
- the correct event name and time,
- required parameters and their data types,
- duplicates,
- the correct property and measurement ID,
- consent status,
- a connection to the actual order or CRM record.
An event visible in DebugView may still not be correctly registered as a custom dimension, available in a standard report or suitable as a key event. Test the whole chain.
Traffic sources need consistent rules
Use a consistent UTM taxonomy for campaigns. Define source, medium, campaign and, where appropriate, campaign ID before launch. Do not tag internal links on your own website with UTM parameters — you would overwrite the original acquisition context.
Payment gateways and external checkouts can incorrectly appear as a new referrer. Before adding a domain to unwanted referrals, check cross-domain measurement, the return flow and the actual cause. A report exception must not hide a broken implementation.
Enable integrations according to decisions
Google Ads
The integration allows selected data, audiences and key events to be used between GA4 and the advertising account. Before importing anything for optimisation, check which event is primary, whether the conversion is duplicated by a Google Ads tag and whether it has sufficient business quality.
Search Console
The integration adds a view of organic queries and landing pages. Search Console and GA4 measure different parts of the journey and use different metrics. The numbers are not meant to be identical.
Merchant Center and other products
An online shop can connect other Google products according to the services and permissions it uses. Do not enable every integration simply because it appears in the administration. Each one adds access, data and control work.
Meta and other advertising systems
The direct Meta connector to GA4 can import aggregate costs, clicks and impressions; it does not transfer Pixel events. The current process and limits are in How to connect Meta Ads with GA4. Server events for Meta are handled separately with Conversions API.
Filter internal traffic in testing first
A company can define internal traffic from available IP ranges and use a data filter. Set it to testing first and verify it in Realtime. An active exclusion filter changes future processing and cannot restore historical data.
For people on mobile networks, dynamic IPs and working from home, an IP filter may not be complete. Document the scope of the limitation instead of creating an illusion of cleanliness. Depending on the situation, internal testers can also be identified through a controlled custom path, provided it does not create a new identity problem.
Retention and export belong in the plan before you need the data
Retention settings affect user and event data available in some analyses. Standard aggregated reports may behave differently. If the company needs a long-term detailed history, its own CRM joins or reproducible analyses, plan the BigQuery export and data management from the beginning.
An export to a warehouse is not “free forever and without work”. It needs a schema, costs, access, retention and someone who recognises a data change. For a smaller website, a good GA4 overview and regular business report may be enough.
Attribution does not create one absolute truth
GA4, Google Ads, Meta and a CRM may assign a different channel to the same order. Each system sees a different signal and uses different attribution conditions. Some processed or modelled values may also change after the event.
Therefore decide:
- which system controls the budget for a particular platform,
- which report compares channels,
- where the actual sale or qualified lead is confirmed,
- which attribution window and model you use to run the company.
A difference between systems is something to explain, not automatically a reason to rewrite one graph until it resembles another.
The minimum for business GA4
- Written business questions and a measurement plan.
- One correct property, data stream and controlled access.
- A verified Google tag or GTM container on all relevant pages.
- Recommended events with required parameters.
- A few real key events instead of dozens of micro-conversions.
- Consent handling tested for both acceptance and rejection.
- Consistent UTMs and corrected payment or cross-domain flows.
- A test order or enquiry compared with the backend.
- Integrations only with services the company uses.
- A simple regular report that someone actually uses to decide.
When GA4 is not enough
GA4 is not a CRM, accounting system or full data warehouse. In a long B2B cycle, you need to connect the marketing entry point with contact quality in sales. In an online shop, the ordering system and accounting confirm the sale. For complex journeys, add your own data layer and export, but only when it is clear who will use them.
Good measurement is not the greatest number of events. It is a state in which people in the company know the limits of the data and can still make calmer decisions from it. If you want to review your current implementation from the business question to the concrete event, describe it via contact.