Google Analytics 4 is not harder than the version before it, but it is different in one way that catches people out: it does not assume anything about what matters to you. Out of the box it records that people visited and how long they stayed. It has no idea that a form submission on one page is worth four thousand dollars and a newsletter signup is worth almost nothing until you tell it.
That is the whole setup job. Install it once, define the handful of things that count as a result, make sure campaign traffic arrives labelled, and build one report you will genuinely open. Everything else in the interface is optional.
Install it once, through a tag manager
Create the property, create a web data stream for your domain, and install the tag through Google Tag Manager rather than pasting code into your theme. The reason is maintenance: everything you add later, including conversion tags, call tracking and third party scripts, goes through the same container instead of accumulating in your site template.
Two mistakes to avoid at install. Do not run two properties collecting the same data as a safety net, because you will end up with two different numbers and no way to decide which is right. And check for an existing tag before you add one, because double tagging inflates every metric and is invisible until somebody notices the bounce rate is impossibly low.
Turn on enhanced measurement, then check what it captured
Enhanced measurement automatically records scrolls, outbound clicks, site search, video engagement and file downloads without any configuration. Leave it on, but do check the results, because automatic tracking picks up things you did not intend. A site search event firing from a filter widget, or a file download event firing from an image link, will quietly distort the reports built on them.
Use the real time and debug views to watch events fire while you click through the site yourself. Ten minutes of this at setup catches problems that otherwise surface three months later when a report stops making sense.
Define key events that match real outcomes
In GA4 the things that used to be called goals are key events, and defining them is the step that turns analytics from a traffic counter into a business tool. For most businesses the list is short: a completed enquiry form, a booking, a phone call over a meaningful duration, a quote request, a purchase.
Two rules keep the list honest. Mark as a key event only what you would be pleased to receive an email about, which keeps soft engagement out of the headline number. And track the confirmation rather than the click, because a button click records intent while a thank you page or a successful submission records the outcome.
Track phone calls or accept that the data is incomplete
For service businesses, a large share of enquiries arrive by phone, and by default none of them appear in analytics. That means the channel report you are using to allocate budget is missing the results of whichever channel drives the most calls, which is usually local search.
At minimum, track clicks on tap to call links as an event, which captures mobile intent. Better, use a call tracking number that records the source and passes it back in, and count only calls over a set duration so that wrong numbers and hangups stay out of the total. Without one of these, comparing channels is guesswork dressed as data.
Filter out your own traffic
Your team visits your site constantly. Staff checking a page, a developer testing a form, an agency reviewing a build. On a site with modest traffic, internal visits can be a noticeable share of sessions and a much larger share of form submissions, because internal test submissions land in the same bucket as real enquiries.
Define internal traffic by IP address in the data stream settings and activate the filter, rather than leaving it in testing mode where it labels but does not exclude. Do the same for any developer or staging domain so pre launch testing never reaches the production property.
Tag every campaign so the source report means something
Traffic that arrives without campaign parameters gets filed under direct or referral, which is where attribution goes to die. Any link you place anywhere outside your own site should carry UTM parameters: the source, the medium and the campaign at minimum.
Agree a naming convention and write it down, because the reports group by exact string. Facebook, facebook and FB are three different sources in the data, and a quarter of mixed capitalisation produces a channel report nobody can total. Keep everything lowercase, use the same word for the same thing every time, and apply it to email links, social bios, printed QR codes and partner placements alike.
Set data retention before you need the history
GA4 defaults to a short retention window for the event level data that powers exploration reports. Change it to the longest setting available in the property admin, and do it now rather than when you first want a year over year comparison, because the setting is not retroactive. Data already discarded does not come back.
While you are in the admin, link the property to Search Console and to your ad accounts. The Search Console link puts the search queries that led to your site into the same interface as the behaviour that followed, which is the connection that makes both datasets more useful.
Understand what your consent banner is doing to the numbers
If your site shows a cookie banner, a share of your visitors decline and are never measured. That share varies by region and by how the banner is written, and it means your reported totals are a sample rather than a census. The practical effect is that your real traffic and real conversion counts are higher than the dashboard says.
This is not a reason to remove the banner where consent is required. It is a reason to compare like with like: judge performance against your own previous periods under the same consent setup, rather than against an absolute benchmark or against a period before the banner existed.
Build one report you will actually read
The default interface offers far more than anyone needs, and the practical result is that most business owners open it twice and never again. Build a single dashboard, in Looker Studio or as a saved collection, with five numbers: key events this month against last, the channels that produced them, the landing pages that produced them, the conversion rate by channel, and the pages losing the most traffic.
Five numbers you look at monthly beat fifty you never open. If a metric has never once changed a decision you made, remove it from the report. The point of measurement is to change behaviour, and a report nobody reads changes nothing.
Read attribution with appropriate suspicion
GA4 will attribute a conversion to a channel and it will do so confidently. Treat the number as directional. Cookie restrictions, cross device journeys, consent banners and the plain fact that people research on a phone and buy on a laptop mean the model is reconstructing a path it cannot fully see.
The correction is cheap: ask every enquiry how they found you and log the answer next to what analytics says. When the two disagree consistently on one channel, believe the humans and adjust how much weight you give the model. Organic search and word of mouth are the two most commonly undercounted.
Connect analytics to what actually happened
Analytics stops at the enquiry. It does not know which enquiries became customers, which is the only question that decides where next quarter's budget goes. Twenty leads from one channel that never close are worth less than five from another that all do, and no analytics property can tell you that on its own.
Closing that loop means joining the enquiry data to your sales records, even if the first version is a spreadsheet with source, date, value and outcome. We build this as a connected dashboard for clients because the answer changes budget decisions more than any other number in marketing.
How Social Signals Marketing sets up measurement
Tracking goes in before the campaign, not after it, because retrofitting measurement means the first month of data is unusable. Key events are defined from the client's actual sales process rather than from a default list, calls are tracked alongside forms, and the monthly report shows cost per enquiry by channel rather than sessions.
See how it supports the work in our paid advertising services, read how to improve your website conversion rate, or contact Social Signals Marketing for an analytics setup review.
Frequently asked questions
Is Google Analytics 4 free?
Yes. The standard version is free and covers everything a small or medium business needs, including unlimited key events, campaign reporting and integrations with Search Console and Google Ads. The paid tier exists for enterprises with very high event volumes, stricter service level requirements or advanced data export needs.
What is a key event in GA4?
A key event is an event you have marked as important, which is what previous versions of Analytics called a goal or conversion. Everything GA4 records is an event; marking a handful of them as key events tells the reports which ones represent a business outcome. For most businesses the list is a completed enquiry form, a booking, a qualifying phone call and a purchase.
Do I need Google Tag Manager to use GA4?
Not strictly, but it is worth using. Without it, every tracking change means editing your site code, and over a couple of years you accumulate a pile of scripts in your template with nobody sure what they do. Tag Manager keeps tags in one place, adds version history, and lets you preview and debug before publishing.
Why does GA4 show different numbers from my ad platform?
Because they count differently. Ad platforms usually credit a conversion to themselves if the person saw or clicked an ad within their attribution window, while GA4 applies its own model across all channels. Neither is wrong; they are answering different questions. Use the ad platform to optimise campaigns and use GA4 to compare channels against each other.
How long does GA4 keep my data?
Aggregated reporting data is retained indefinitely, but the event level data that powers exploration reports defaults to a short window and can be extended in the property settings. Change it to the maximum as soon as the property is created, because the change is not retroactive and data outside the window cannot be recovered.