/GOOGLE ANALYTICS 4
GOOGLE ANALYTICS 4
SET UP SO THE NUMBERS MEAN SOMETHING
GA4 will report numbers on the day you install it. Whether those numbers describe your business is a separate question, and usually the answer is no. This page covers the model it uses, the order to configure it in, and the misconfigurations we are asked to fix most often.
/THE MODEL
IS DIFFERENT
The central change from the previous generation of Analytics is that everything is an event. A page view is an event. A purchase is an event. A scroll is an event. There is no separate hierarchy of pageviews, events with categories and labels, and sessions built on top of them — there are events with parameters, and sessions are derived from them.
This is more flexible and considerably less forgiving. Nothing stops you from sending a hundred event types with inconsistent names, and nothing will warn you that half of them are unusable six months later when you need to build a report. A naming convention decided before implementation is worth more than any amount of configuration afterward.
The second consequence is that metric definitions changed. Numbers that look like their old equivalents are frequently measuring something else, which is why year-on-year comparisons across a migration are a reliable source of unnecessary alarm.
/THE SETUP
IN ORDER
Each step depends on the ones before it. Doing them out of order is how accounts end up with data that cannot be repaired retroactively.
Property and data stream
One property per business unit you actually report on separately, not one per site out of habit. Set the data stream up before anything else and decide the time zone and currency now — changing them later does not restate historical data.
A tag management layer
Deploy through Tag Manager rather than hard-coding, unless you have a specific reason not to. It costs an hour at the start and saves a development ticket every time a measurement requirement changes, which is often.
Events that map to business actions
Decide what actually matters — a submitted enquiry, a completed purchase, a booked call — and name those events deliberately and consistently before implementing any of them. Automatic enhanced measurement gives you scrolls and outbound clicks for free; it does not give you anything specific to your business.
Mark the ones that count
Only the events that represent real business outcomes should be marked as key events. Marking everything makes the conversion figure meaningless and, if the property is linked to an advertising account, actively teaches automated bidding to optimize toward noise.
Link the other Google properties
Search Console for organic query data, and Google Ads if you are running paid. The links are what let you see the whole path in one place; without them you are reconciling exports by hand.
Consent, retention and internal traffic
Configure consent handling properly, set data retention deliberately rather than accepting the default, and exclude internal and staging traffic. This last one sounds trivial and is responsible for a great many puzzling conversion counts.
Verify before you trust anything
Use preview and real-time reporting to confirm each event fires once, on the right trigger, with the right parameters. Then wait for full processing before comparing anything to another system.
/WHAT WE FIX
MOST OFTEN
In rough order of how frequently we find them on accounts brought to us for a second opinion.
- Everything marked as a conversion. Newsletter signups, PDF downloads and scroll depth counted alongside purchases, producing a conversion number that no one in the business recognizes.
- Double tagging. The tag deployed both directly in the theme and through Tag Manager, inflating every count and halving every rate, often for years before anyone notices.
- Internal traffic never excluded. The agency, the client’s own office and a monitoring service all counted as customers — most visible on low-traffic B2B sites, where it can be a large proportion of the total.
- Thank-you page conversions that fire on refresh. A conversion triggered by a page view rather than a transaction, counting again every time somebody reloads or returns to the page.
- Cross-domain measurement not configured. A checkout or booking system on a different domain, so every purchase is attributed to a referral from your own site.
- Consent implemented as a banner and nothing else. The banner is displayed, the tags fire regardless, and the business is both non-compliant and confident in its data.
- Data retention left at the default. Discovered at the moment somebody needs a year-on-year comparison and finds the data no longer exists.
/FREQUENTLY ASKED
QUESTIONS
NOT SURE YOUR
DATA IS TELLING THE TRUTH?
A measurement review is a couple of days of work and it usually changes at least one decision the business was about to make. Tell us what you have installed and what you are trying to measure, and we will tell you what we would check first.
- Conversion tracking — the advertising side of the same problem
- Data and AI — when the reporting needs to leave the platform
- Page experience — the other set of numbers people misread
- All Insights — the full index of guides, tools and updates
Start your journey
Tell us what you have installed and what you need to measure.