Setting Up GA4 So the Numbers Mean Something
A default GA4 install collects data immediately, which creates the impression that it is set up. It is not. Without a handful of specific configuration steps it will overstate traffic, misattribute revenue, and quietly discard the history you will want in a year.
Fix the settings that silently lose data
Start with data retention. The default keeps user-level and event-level detail for a short period, after which exploration reports can no longer reach back. Extending it to the maximum available costs nothing and cannot be applied retrospectively, which is exactly why it should be the first thing you change.
Next, internal traffic. Your own team browsing the site inflates sessions and pollutes behaviour data, and on a low-traffic site it can dominate. Define internal traffic by IP and activate the filter, remembering that defining it is not the same as switching it on.
Then referral exclusions. Payment gateways are the classic problem: a customer leaves to pay, returns, and the return is recorded as a new visit from the gateway, which then takes credit for the sale. Add your payment providers and any authentication domain to the unwanted referrals list, or your revenue reporting will credit the wrong sources indefinitely.
Define what actually counts as success
The default events tell you almost nothing about your business. Page views, scrolls and outbound clicks are activity, not outcomes. You need events for the things that matter: a form submitted, a call started, a booking made, a purchase completed.
Mark only genuine outcomes as key events. Marking a newsletter signup, a scroll and a video play all as conversions produces a conversion count that no longer means anything, and any optimisation built on it will chase the cheapest of them.
Then verify the events actually fire, using the real-time and debug views, by completing the action yourself on a phone as well as a desktop. A surprising share of tracking failures are mobile-only, and nobody notices because the desktop test passed.
Tag your campaigns or lose the attribution
Traffic from email, social posts, QR codes and paid placements arrives untagged as direct or referral, and once it does the information is gone. Consistent campaign tagging is what makes the acquisition reports usable, and it has to be applied by everyone, every time.
Consistency is the hard part. Facebook, facebook and FB are three different sources in the reports. Agree a lower-case convention, write it down, and generate links from one place. Our UTM builder exists to remove the manual typing that causes this.
Link GA4 to Search Console and to your ad accounts. The Search Console link brings query data into the same place as behaviour data, and the ads link is what allows imported conversions and proper cost reporting rather than two disconnected pictures.
Know what the numbers cannot tell you
Analytics undercounts. Ad blockers, tracking protection in some browsers, and users declining consent all remove data, and the loss is not evenly distributed: technical audiences are blocked far more than general ones. Treat the totals as a consistent index for spotting trends rather than an exact count.
GA4 also applies data thresholds that suppress small numbers to protect identity, and it samples large explorations. If a report shows fewer rows than you expect, that is often why rather than a tracking fault.
Above all, reconcile against something real. Compare GA4 purchases against the orders in your own system, and leads against what actually arrived in the inbox or CRM. A consistent gap is tolerable once you know its size; an unknown gap makes every decision built on the data a guess. Our data analytics service starts with that reconciliation.
Key takeaways
- Extend data retention first. It cannot be applied retrospectively, so every day at the default setting is history you lose.
- Define and activate the internal traffic filter, and add payment gateways to unwanted referrals or they will steal credit for sales.
- Mark only real business outcomes as key events. A conversion count that includes scrolls is meaningless.
- Test that events fire on a phone as well as a desktop; many tracking failures are mobile-only.
- Reconcile GA4 against your own order or CRM data. A known gap is workable, an unknown one is not.
Frequently asked questions
Why does GA4 show different numbers from my ad platform?
Because they count differently and neither is wrong. Ad platforms typically credit a conversion to the click that preceded it within their own attribution window and count conversions on the day of the click, while GA4 attributes across all channels using its own model and counts on the day of the conversion. Ad platforms also see conversions from users who blocked analytics. Expect a persistent difference, learn its usual size, and reconcile both against your actual order data rather than trying to make them agree.
What should I mark as a conversion in GA4?
Only outcomes that represent real business value: a purchase, a qualified enquiry, a booking, a call started, a demo requested. Marking secondary engagement such as scroll depth, video plays or newsletter signups as key events makes the conversion total meaningless and, worse, any automated bidding you feed with it will optimise for whichever is cheapest to produce. If you want to track those softer actions, keep them as ordinary events you can still report on without polluting the main figure.
Why is so much of my traffic showing as direct?
Direct is the fallback bucket for traffic GA4 cannot attribute. It includes genuine direct visits, but it also collects untagged links from email and messaging apps, clicks from documents and desktop applications, and traffic that lost its referrer moving from a secure to an insecure page. The fix is consistent campaign tagging on every link you control. If direct is unusually large and growing, check whether an email or social campaign is going out untagged.
How long does GA4 keep my data?
Event and user level data is kept for a limited default period that is shorter than most people expect, after which detailed exploration reports can no longer reach further back. Standard aggregated reports are unaffected. Extending retention to the maximum available in your property settings takes seconds and should be done immediately, because the change is not retrospective: any detail already discarded is gone. If you need genuinely long history, export raw data to a warehouse.
Do I need Google Tag Manager as well as GA4?
Not strictly, but it helps once tracking becomes more than a page view tag. Tag Manager lets you add and change tracking without a developer deploying code each time, keeps all your tracking in one auditable place, and makes it far easier to control which tags fire under which consent conditions. For a simple site with one analytics tag and no advertising pixels, installing GA4 directly is perfectly reasonable and one less thing to maintain.
How accurate is GA4 data?
Directionally reliable and absolutely incomplete. Ad blockers, browser tracking protection and consent refusals all remove sessions, and the loss is uneven because technical and privacy-aware audiences are blocked far more than general ones. GA4 additionally suppresses small numbers to protect identity and samples very large queries. Use it to compare periods, channels and pages against each other, and use your order system or CRM for anything where the absolute number matters.