GA4 audit checklist
Twenty GA4 checks in four groups, tickable on this page with progress saved locally, plus a printable PDF and an Excel measurement plan that starts from the business question rather than the event. Check one is that the tag fires exactly once: a double-fired conversion makes every downstream decision wrong, and nothing on the surface looks broken.
- CHECKLIST
- XLSX
- FREE
- Best for
- Anyone who inherited a GA4 property, or set one up and never checked it
- Includes
- 20 tickable checks + Excel measurement plan + printable PDF
- Time to use
- 90 minutes
Free to download and use in your own client work. No email address required.
Created by Alston Antony Measurement tied to signups and demos, not sessions Updated
Problems this solves
Recurring complaints about this task, and what this resource does about each one. If it does not solve your version of the problem, that is worth knowing before you download anything.
- Conversions are counted twice and every report still looks normal
- Check one is verifying the tag fires exactly once in DebugView, and the conversions group names the specific cause people miss: importing the same conversion from GA4 AND from the Ads tag, so two Primary actions count one sale.
- The conversion tag fires on a page view rather than on the actual outcome
- The checklist requires each key event to be triggered by you and verified, and requires form submissions to be measured on the confirmation rather than the button — which otherwise counts everyone who hit a validation error.
- You want a year-on-year comparison and the data is not there
- Data retention defaults to two months and cannot be recovered retrospectively. The checklist flags it, and the page says plainly it is the highest-regret default in GA4.
Analytics is the audit to run first, because every other channel audit compares performance against this data. If it is wrong, you are auditing noise.
And the most common defect is not a subtle one. The tag fires twice.
The checklist
Installation 5
Conversions 7
Data quality 6
Usability 4
Three settings almost everyone leaves on the default
Data retention is 2 months out of the box. Change it to 14. You cannot recover data you did not retain, and any year-on-year comparison needs more than two months of history. This is the single highest-regret default in GA4.
Unwanted referrals is empty. Without it, your own payment provider gets credited for the sale, and the referral report becomes fiction — which matters most for exactly the sites that make money.
Internal traffic is not filtered. Your own visits inflate the pages you look at most, which is the worst possible direction for a bias to run.
Verify the tag fires once, in DebugView
Not by assumption, and not by checking the page source — a tag added both in the site template and in Tag Manager renders once in the HTML and still fires twice. Open DebugView or preview mode, load a page, and count the events. It takes two minutes and it is the highest-value check on this page.
The two causes of double-counted conversions
Double counting is the most common analytics defect there is, and it has two causes worth separating because the fix is different.
A tag added in two places. The site template and Tag Manager both load GA4, so every pageview and every event fires twice. Visible only in DebugView or the browser console — the page source looks correct.
The same conversion imported twice. This is the one people miss. The conversion arrives in Google Ads once from the GA4 import and once from the Ads conversion tag on the page, so two separate conversion actions each count the same sale. In Google Ads it looks like genuinely doubled performance.
The fix for the second is to decide which source you trust and set only that one to Primary, moving the duplicate to Secondary. Secondary actions still report but do not feed bidding, which is the part that matters — a bid strategy optimising against doubled conversions will keep raising bids on the strength of sales that did not happen.
How to tell which one you have
If GA4 and Google Ads both show roughly double your real sales, it is the tag in two places. If GA4 looks correct and only Google Ads is inflated, it is the import plus the tag. That one distinction saves an afternoon of looking in the wrong tool.
The measurement plan
Most GA4 setups fail because they start from the event — someone tracks every click and then tries to build a report from it. Start from the question instead.
| Column | What to put in it |
|---|---|
| Business question | In plain words: "which channel produces enquiries that close?" |
| Metric | The single number that answers it. |
| Event or source | Where the number comes from — GA4 event, Search Console, CRM. |
| Definition | Written down so two people calculate it the same way in six months. |
| Owner | Who is responsible for it being correct. |
| Reported where | Weekly, monthly report, dashboard, or nowhere yet. |
| Status | Working, partial, or not built. Dropdown. Most plans start mostly "not built". |
The definition column is the one that saves arguments. Written down, “qualified enquiry” means the same thing to two people in six months’ time; undefined, it means whatever each person assumed when they built their own version of the report.
The workbook arrives with four questions filled in, because a blank measurement plan is the hardest kind to start. It also counts how many of your questions have nothing built yet — which, honestly, is the useful output on a first pass. Most plans start out mostly “not built”, and knowing that is the finding.
Common mistakes
- Reporting from GA4 before verifying the tag fires once.
- Leaving data retention at two months, then wanting a year-on-year comparison.
- Measuring form submissions on the button rather than the confirmation.
- Marking a dozen events as key events, so none of them means anything.
- Comparing GA4 sessions to another tool’s sessions as if they counted the same thing.
- Building events first and hoping a report emerges from them later.