Google Analytics 4 shows hundreds of numbers, and most of them help you decide nothing. If you open the dashboard once a month, see that sessions went up and close it, GA4 is working as decoration.
This article separates what helps you decide from what merely fills the screen, and shows how to set up the useful minimum.
Why sessions and pageviews mislead
Sessions and pageviews are easy to raise and hard to interpret. A shared post, a spike from social media or automated traffic inflate the number without producing a single client.
The problem is not that the data is false. It is that it does not answer the question you have. The real question is almost never "how many people came". It is "how many people did something that matters, and where did they come from".
There is one case where sessions do matter: comparing against yourself over time, within the same source. If organic traffic dropped 40% in a month, that signals a technical problem or lost rankings, and it is worth investigating with the technical SEO checklist. Beyond that, sessions alone are noise.
The GA4 event model (and why it differs from Universal Analytics)
This is the conceptual shift that confuses anyone arriving from the previous version. In Universal Analytics the world was made of sessions, and the pageview was the central unit. Events existed on the side, with category, action and label.
In GA4, everything is an event. A pageview is the page_view event. The start of a session is session_start. A click leading off the site is click. Every event carries parameters, which are the details of that occurrence.
Two things change in practice. First: you no longer force reality into category and action; you name the event for what it is and add the parameters that matter. Second: enhanced measurement already collects a set of events automatically when the property is set up, including scrolls, outbound clicks, site search and file downloads. Worth opening the data stream settings and seeing what is already on before installing anything.
There are three families of events: automatically collected ones, the ones Google recommends (standard names like generate_lead or purchase, which unlock ready-made reports) and custom ones you invent. Whenever a recommended name exists for what you want to measure, use the recommended name.
Engagement rate: what replaced bounce rate
Universal Analytics measured bounce rate as sessions with a single interaction, which unfairly punished pages that answered the visitor question well. Someone who arrived, read the opening hours and left satisfied counted as a bounce.
GA4 flipped the logic. It defines an engaged session as one lasting more than 10 seconds, or firing a conversion event, or containing at least two pageviews. Engagement rate is the share of engaged sessions out of the total. Bounce rate came back later, but now it is simply the complement: everything not engaged.
Two practical consequences. The 10-second threshold is adjustable in the property settings, and raising it makes sense for long-form content. And, more importantly, engagement rate is useful for comparing pages against each other, not as an absolute grade. If one service page sits far below the others, the cause is usually a broken expectation: the ad or the search result promised something the page does not deliver.
Also worth watching is average engagement time, which in GA4 counts only time with the tab in the foreground. It is a more honest measure than the old session duration, which counted forgotten open tabs.
Conversions: marking your first one in ten minutes
With no conversion marked, no other report means anything. It is the most important GA4 setting, and the most postponed.
The fastest route, without touching code, uses the interface itself. First, perform the action you want to measure on the site, for instance submitting the contact form, and confirm in the realtime report that some event appears. If the submission leads to a thank-you page with its own URL, that page_view already works.
Then, in Admin, use event creation to build a new event out of the existing one, with a condition. Example: create the generate_lead event when page_view has page_location containing /thank-you. Finally, in the events list, mark it as a key event, which is what GA4 now calls what used to be called a conversion.
If the form does not redirect to a thank-you page, you will need to fire the event on successful submission, usually through Google Tag Manager or a line in the site code. In that case fire on request success, never on the button click: a click is not a submission, and a failed submission is not a lead.
Start with a single conversion, the one that represents money. Contact submitted, phone number tapped, quote requested. Five conversions marked on day one means none of them watched by month two. And if messaging is your main channel, the click has to become an event too, covered in contact form or WhatsApp.
Source and medium: finding where real clients come from
With the conversion marked, the acquisition report finally answers the right question. Open traffic acquisition, group by source and medium, and compare columns: sessions on one side, key events on the other.
It is common for the source with the most sessions not to be the one with the most conversions. Social tends to bring volume and little intent; organic search brings fewer people and more of them ready to buy; direct traffic mixes people who already know you with whatever the tool could not identify.
To know where each campaign visit came from, tag your links with UTM parameters. Three are enough to start: utm_source (where the link was published), utm_medium (the type, such as cpc, social, email) and utm_campaign (the name of the push). Pick a writing convention and stick to it: all lowercase, no spaces. A source written Instagram and another written instagram become two separate rows and you lose the comparison.
One caution: never put UTM parameters on internal links within your own site. That ends the current session and restarts attribution, making your own site appear as the source of your conversions.
A monthly report that fits on one page
A report nobody reads does not exist. The format that survives fits on one page and answers five questions.
| Question | Where to look in GA4 |
|---|---|
| How many enquiries arrived? | Key events for the period |
| Where did they come from? | Traffic acquisition, by source and medium |
| Which pages produced contacts? | Pages and screens, with the key events column |
| Did anything drop sharply? | Comparison with the previous month |
| What are we changing this month? | One sentence, written by you |
The last line is what turns data into a decision. Without it, the report is a bank statement nobody uses to plan anything.
Two habits help: always compare with the same month a year earlier when the business is seasonal, and note in a calendar whenever something changed on the site. An unexplained drop usually has a date, and that date usually matches an edit, a redesign or a performance regression.
Consent Mode and the data you will not have (and why that is fine)
If your site has a cookie banner, and it should if you use analytics or advertising tools, some visitors will decline. Without consent, identifiers are not stored, and those visits do not enter the reports in the usual way.
Consent Mode is the Google mechanism for handling that: tags receive the consent state and adjust what they send. With consent denied, instead of sending nothing they send signals without identifiers, and Google uses modeling to estimate the behavior it could not measure. It matters to understand what that means: part of your numbers becomes an estimate, not a count.
Two practical consequences. First, the conversion total in GA4 may not match the number of form emails you received, and that is not a bug. Second, comparing months remains valid, as long as the consent setup did not change midway.
The obligation to ask for consent, the wording of the banner and the right to decline without penalty are covered in the article on privacy law for small websites. Read it before configuring Consent Mode, because the legal decision comes first and the technical one after.
And it is worth staying calm about it: an approximate number about where your clients come from still decides better than an exact number about pageviews.
If you want GA4 set up with a real conversion tracked and a report someone actually reads, talk to ALB Seven. It is a short configuration, and it changes every decision that follows.