Five quick GA4 checks catch most errors that distort the reading of ads: duplicated tags, badly defined conversions, misclassified sources, internal traffic included and misconfigured consent. Doing them quarterly is cheap insurance.
The essentials in a definition
A GA4 audit with five checks catches most errors: correct key events with no duplication, consent working, internal-traffic and spam filters, cross-domain linking and matching with the database. It builds on key events, the technical documentation and consent configuration.
1. Tags and events
Confirm the GA4 tag fires once per page, without implementation duplications (manual code and tag manager at the same time). Check the real-time view. Test the main events. Duplications inflate sessions and conversions.
One tag, one firing.
2. Conversions
Confirm that only important actions are marked as conversions, with correct values and currency. Compare the number of purchases or leads with the sales system. Check that the transaction ID avoids duplicates. Moderate differences are normal; large differences indicate error.
Wrong conversions contaminate all optimisation.
Five checks that catch most GA4 errors
- Tags without duplication
- Conversions with correct values
- Well-classified sources
- Internal traffic excluded and consent correct
3. Sources and campaigns
See whether a large part of traffic appears as “unassigned” or “direct”. Check campaign tags and the Google Ads link. Confirm referral exclusions for payment and third parties. Misclassified sources hide channel performance.
Without correct source, there is no attribution.
Five checks a quarter avoid months of decisions based on wrong data.
4 and 5. Internal traffic and consent
Exclude internal traffic (team, agencies, tests) by IP or another method. Check that the consent banner and consent mode work and that tags respect choices. Test different scenarios (accept, refuse) and devices. Document and repeat quarterly.
Clean data starts by excluding internal noise.
Decision table
| Check | Question | How to test |
|---|---|---|
| Key events | Are they correct and without duplication? | Debugging and comparison with orders |
| Consent | Do tags respect the state? | Test accept, decline and change |
| Internal traffic and spam | Is it filtered? | Review filters and reports |
| Domains and sessions | Does the session persist across domains? | Walk through the complete flow |
Common mistakes to avoid
- Auditing only the interface. Implementation errors go unseen.
- Not comparing with the database. References are lacking.
- Fixing without documenting. History is lost.
How to measure the result
Track the number of problems found, correction time and agreement with the database.
Limits and exceptions
Five checks catch most, not all errors. The interface and options change; confirm current documentation. Deeper audits may be necessary in complex implementations.
Illustrative example
Illustrative example: a team does the five checks in a morning. It finds a purchase counted twice, unfiltered internal traffic and a tag firing before consent. It fixes in order of impact and documents.
In practice: checkpoints
Prepare a sheet with the five checks, the test method for each and the date of the last run, and share it with whoever manages the site and campaigns. For key events, make a test purchase and a test lead and compare them with the database. For consent, test the three scenarios and confirm that tags respect the state. For filters, check that internal traffic and spam are excluded and that retention settings are right. For domains, walk through the complete flow from ad to purchase. Finally, compare totals with the database. Fix by order of impact, record each change and put the next audit in the calendar before closing this one. Keep screenshots of the settings before and after, so you can show what changed, and store the findings where management can read them. Add the audit to the release process, so that a new theme, plugin or checkout step triggers a quick re-check. Many serious measurement problems stay hidden for months simply because nobody looks, and a short, regular audit is the cheapest way to avoid paying for decisions based on faulty data.
Frequently asked questions
How long does it take?
One to two hours, if everything is documented.
What is the most common error?
Duplicated tags or conversions.
Who should do it?
Whoever manages measurement, with cross-review.
How often should I audit?
Quarterly and after important changes.
Who should audit?
Whoever knows the implementation and, preferably, a second pair of eyes.
Run five regular GA4 checks (key events, consent, filters, domains and database) and document the fixes.
Sources and recommended reading
Last editorial review: October 2026. Informational content; always confirm current conditions in the official sources before deciding.