Optimisation starts with confidence in your data

Installing GA4 is not the same as measuring reliably. A report may show visitors, revenue and conversions while important events are missing, purchases are duplicated or a payment provider replaces the original traffic source. Acting on those numbers can send optimisation effort to the wrong part of the journey.

A useful review follows one complete order from landing page to confirmation. It compares what the customer actually does with what GA4, the ecommerce platform and marketing tools record. These eight checks help you identify which data is dependable, where uncertainty remains and what needs fixing first.

1. Define which decisions GA4 must support

Start with the questions you need to answer, not with every report available. Which channels attract relevant visitors? Which product pages lead to cart activity? Where does checkout lose momentum? Which campaigns support new customers or repeat purchases?

Connect each question to a small set of events and dimensions. Product views, add-to-cart actions, checkout steps, purchases, revenue, currency, campaign source and market or language are usually more valuable than dozens of interactions that never change a decision.

2. Test the complete ecommerce sequence in DebugView

Google recommends standard ecommerce events such as view_item, add_to_cart, begin_checkout and purchase with the relevant parameters. Test them in DebugView while completing a realistic order. Check not only whether the event appears, but whether product ID, name, price, quantity, currency, coupon, transaction ID and order value make sense.

Repeat the test on mobile and with useful variations: without a discount, with a discount, with several items and — where relevant — another language or currency. Reports and explorations may take time to populate; DebugView lets you inspect the implementation during the test.

3. Check purchases for missing and duplicate transactions

Compare a representative period in GA4 with the ecommerce or payment system. Totals may differ because of consent, refunds, time zones and definitions, but large or unexplained gaps deserve investigation.

Use a unique transaction ID for every purchase and verify that refreshing a confirmation page does not send the same purchase again. Review cancellations, test orders and refunds too. Higher measured revenue is not good news when one order has been recorded twice.

4. Protect the original traffic source

When customers leave for another domain during payment, their return can be treated as a new referral. The payment provider may then receive credit for a sale that originated from organic search, email or a campaign.

Review unwanted referrals and configure cross-domain measurement when the journey genuinely spans several domains you control. Sudden increases in direct traffic or referrals from your own domains can also indicate that sessions or campaign context are being interrupted.

5. Test campaigns with consistent UTM rules

Create one naming convention for source, medium and campaign, then apply it across email, social, partnerships and other channels. Variants such as newsletter, email and e-mail split one channel unnecessarily across reports.

Open a real test email or campaign link and confirm that its parameters survive redirects and remain connected to the important events. A perfectly written UTM tag is useless when the next step strips the parameters or starts a new session.

6. Review consent without disguising missing data

Consent mode communicates visitors' consent status to Google tags; it does not provide a banner or replace legal advice. Test separately what happens before a choice, after acceptance and after refusal. Confirm that the signals being sent match the behaviour of the consent interface.

Fewer observed users after a correct privacy implementation does not automatically mean tracking is broken. Under qualifying conditions, GA4 may show modelled data, so reports can combine observed and modelled behaviour. Document the consent mode, reporting identity and privacy configuration before comparing periods.

7. Remove internal traffic and duplicate measurement layers

Employees, agencies and developers may visit product pages, forms and checkouts far more often than customers. Identify internal traffic and test a filter before activating it permanently; active data filters work forward and do not repair historical data.

Check whether the same GA4 property is loaded through several routes, such as an app, Google Tag Manager and a hardcoded script. Shopify manages pixels and customer events in Customer events. Keep a record of connected pixels, their privacy requirements and the owner of each implementation.

8. Turn the review into a measurement plan and changelog

For every important event, document the trigger, parameters, platform, owner and latest test date. Note which events are marked as key events and which reports the team actually uses. This shows what needs retesting after a redesign, checkout change, consent update or market launch.

Finish with a short priority list: missing or duplicate purchases first, broken attribution and consent second, followed by product parameters, campaign consistency and reporting. Conversion optimisation and channel comparisons become meaningful only when that foundation is sufficiently reliable.

A dashboard is not proof — a tested journey is

Analytics creates value through the decisions it supports, not through the number of charts it contains. Regularly test one complete customer journey and compare systems. Do not expect perfect equality; look for explainable differences and consistent definitions.

This turns GA4 from a separate reporting screen into a working part of marketing. You know which figures are strong enough to act on, which limitations need to be stated and which improvement deserves attention first.

Sources