STERLING LAB
SINCE2013
MARKETS10 COUNTRIES
AUDITFREE · 5 DAYS

GA4 Setup Checker

Short answer

The GA4 Setup Checker is a free self-audit of whether you can rely on your GA4 numbers. You answer 14 yes, no or not-sure questions and get a score out of 100 and a list of likely gaps, biggest first, each with a link to where to read more. It connects to nothing and reads nothing from GA4, and your individual answers are not sent anywhere.

Start the check

Updated · markdown · JSON spec

Run the self-check

Events and conversions

0 of 4 answered
You have written down, in plain words, what a real conversion is for your business (a submitted quote form, a call, a paid order), and the key events marked in GA4 match that list.

GA4 Admin → Events (or Key events): write what each marked event means in real life, then compare with your list. A button click, a scroll or a thank-you page view counted next to a real lead is a mismatch.

One form submission or one order produces exactly one event in GA4, not two or more.

Submit your own test form or place a test order, then open GA4 Realtime or DebugView and count the events that arrive for it.

After the last change to the site, forms or tags, someone placed a test order or submitted a test form and watched it arrive in GA4 under the right event and source.

GA4 → Realtime (or Admin → DebugView): complete a conversion on your site and look for the event and the source it is credited to.

Purchase events carry the items, the currency, the value and a transaction ID, so an order is not counted twice and revenue in GA4 can be compared with your orders.

Open a test purchase in DebugView and check the event parameters: items, currency, value and transaction_id should all be present.

Traffic and sources

0 of 5 answered
You have looked at how much traffic GA4 credits to Direct or Unassigned, and you know the main reasons it is there.

GA4 → Reports → Traffic acquisition: look at the Direct and Unassigned rows next to your active campaigns. Typical causes: untagged ad or email links, a payment or booking page on another domain, consent hiding visitors until they accept.

Paid, email and social links carry UTM tags, and one written convention (names for source, medium and campaign) is used across platforms.

Open the links in your live ads and email tools: check for utm_source, utm_medium and utm_campaign, and compare the spelling with the convention. Auto-tagging stays on in Google Ads.

Your channel grouping puts Meta traffic under Paid Social and Google ads under Paid Search, and brand-name organic is not shown as paid.

GA4 → Traffic acquisition with the default channel group: check which channel your Meta campaigns and Google campaigns appear in. If they land under Direct or Organic, the grouping does not understand your UTM convention.

When a visitor moves to a payment page, a booking tool or another domain of yours and comes back, GA4 keeps them in one session instead of starting a new one.

Make a test order that leaves your domain: after the return, check in Realtime or DebugView whether the source of the conversion is your campaign or Direct.

After a visitor accepts the cookie banner, Google’s tags receive a consent update, so GA4 is not left in the default denied state.

Access and data quality

0 of 3 answered
The GA4 property belongs to your business, and someone on your side has admin access, not only a former agency or freelancer.

GA4 Admin → Property access management: your business login should be listed as Administrator. If only an outside login is listed, ask for access before anything else.

Visits by your team and your own testing are filtered out of the reports.

GA4 Admin → Data collection and modification → Data streams → Configure tag settings → Define internal traffic. Then check that the filter is active, not only defined.

GA4 is linked to Google Ads, and the conversions imported into Ads are the same events you trust in GA4.

Checks against reality

0 of 2 answered
In the last month someone compared the number of leads or orders in GA4 with a source outside it, such as your CRM, call log, booking inbox or backend.

Take the same period in GA4 and in your own records and compare the counts. Close numbers mean GA4 is usable. A large gap means you should not spend against GA4 until you know why.

GA4 is checked again before a significant budget change and after a site redesign, a form change or a new booking tool.

Write down the date of the last full check. The short version: ownership, key events, a test conversion, the source it was credited to, a comparison with your records.

What the checker looks at

The questions follow GA4 from what it counts to whether you have checked it against anything outside it. Each one has a short note on where to look in GA4. Events and conversions covers whether your key events match a written list of real conversions, whether one submission gives one event, whether a test conversion was seen arriving and whether purchase events carry items, currency, value and a transaction ID. Traffic and sources covers Direct and Unassigned traffic, a documented UTM convention, channel groups that route Meta and Google correctly, a visitor who keeps one identity across other domains and the cookie banner’s consent update. Access and data quality covers who owns the property, whether your own visits are filtered out and whether Google Ads is linked. Checks against reality covers a comparison with your CRM, calls or orders, and a re-check after changes.

How the score is calculated

Each check has a weight of 1, 2 or 3. Checks that decide whether the numbers in GA4 can be trusted at all weigh the most. A “Yes” earns the full weight, “Not sure” earns half, “No” earns nothing. A check marked as not applicable is left out of both sides of the sum.

The checks with weight 3 are core checks: Key events match your written list of real conversions; One submission, one event; The property is yours and you have admin access. A “No” or “Not sure” on any of them keeps the result out of the top band, even when the score is 80 or higher, and the result names the core check to look at first.

The list of gaps is sorted by points lost, biggest first: a “No” loses the full weight of the check, a “Not sure” loses half of it. Checks that lose the same number of points keep the order of the table below.

Score = (sum of weights answered Yes + ½ × sum of weights answered Not sure) ÷ sum of weights of the applicable checks × 100, rounded to a whole number.

All 14 checks and their weights, 27 points in total
CheckStageWeight
Key events match your written list of real conversionsEvents and conversions3
One submission, one eventEvents and conversions3
A test conversion seen arrivingEvents and conversions2
Purchase carries items, currency, value and a transaction IDEvents and conversions2
Direct and Unassigned traffic checkedTraffic and sources2
One documented UTM conventionTraffic and sources2
Channel groups route Meta and Google correctlyTraffic and sources1
The visitor keeps one identity across checkout and booking domainsTraffic and sources2
The cookie banner passes consent to GoogleTraffic and sources1
The property is yours and you have admin accessAccess and data quality3
Your own visits are filtered outAccess and data quality1
Google Ads is linked and imports the events you trustAccess and data quality2
GA4 compared with your CRM, calls or ordersChecks against reality2
Re-checked after changes and before budget decisionsChecks against reality1

The checks draw on our GA4 audit guide, our GA4 setup service page and our Consent Mode v2 article. Each gap in your result links to the page it comes from; the pages are also listed under Read more.

Worked example

A business that sells nothing online marks the purchase check as not applicable (“No online sales”), which leaves 25 of the 27 points in play. A test form submission arrives twice in GA4, so “One submission, one event” is No. The agency that set GA4 up left no notes, and nobody on the team can say whether a UTM convention, the Google Ads link and the internal-traffic filter are in place, so those three are Not sure. Everything else is Yes.

Use Not sure only when you really cannot tell. If you know a check is missing, answer No.

  1. Applicable weight: 27 − 2 = 25
  2. Lost: 3 + 1 + 1 + 0.5 = 5.5, so earned is 19.5
  3. Score: 19.5 ÷ 25 × 100 = 78

The list starts with “One submission, one event”, then the UTM convention, the Google Ads link and the internal-traffic filter. “One submission, one event” is a core check, so the result names it as the first thing to look at. The band is 50 to 79.

How to read your result

80 to 100
Solid on paper. The remaining items are tidy-up work. Re-run the check before a budget change and after a site, form or tag change.
50 to 79
Gaps worth fixing. Some numbers in GA4 may be counted twice, credited to the wrong source or never checked. A failed or unconfirmed core check holds you here even with a score of 80 or more. Start with the first item in the list.
0 to 49
Hard to trust. Fix the top items before you change budgets, then run the check again.

These bands are our own cut-offs for deciding what to do next, not a benchmark. The order of the list matters more than the number itself.

What a self-check cannot see

Answers describe what you believe is set up. The checker cannot see the property, so it cannot tell a setting that is defined from one that is active, or an event that arrives on one page and not on another. Two checks are the real test: a conversion you place yourself and watch arrive in GA4, and a comparison of GA4’s lead or order count with your CRM, call log or booking inbox.

The neighbouring checkers cover Google Ads and GA4 as a whole, the Meta Pixel with the Conversions API, and the consent chain. They are listed under Read more.

Questions about the checker

Does this connect to my Google Analytics account?

No. It asks questions and scores your answers. Nothing is loaded from your GA4 property, Google Tag Manager, Google Ads or your site, so the result is only as good as the answers. Each question has a short note on where to look in GA4: Admin, Realtime, DebugView or Traffic acquisition.

Why does “Not sure” count as half?

An unknown is not a confirmed failure, but it is not a pass either: nobody is watching that check. Half the weight keeps it visible in the score and on your list. Use it only when you really cannot tell. If you know a check is missing, answer No.

Is the score a benchmark against other properties?

No. The weights and the 80 and 50 thresholds are our own definitions, chosen to rank what matters most for whether you can rely on GA4 numbers. They are not drawn from industry data, and the score does not compare you with anyone.

What happens to my answers?

They stay in your browser tab, are not sent anywhere, and disappear when you leave or start over. Our site analytics receives only an aggregate: that the check was started and finished, with the final score, the band and the number of gaps, and the score again if you click the service link. The individual answers are never sent.

Why can a high score still stay in the middle band?

The three checks with weight 3 are: key events match your written list of real conversions, one submission produces one event, and the property belongs to your business with admin access on your side. A No or a Not sure on any of them keeps the result out of the top band, even when the score is 80 or higher, because the rest of the setup cannot make up for numbers you cannot trust at the base. This rule is our own definition, like the weights and thresholds.

Where do the questions come from?

From our published guide “A GA4 Audit You Can Finish in One Afternoon”, from the description of what we set up on our GA4 setup service page, and from our article on Consent Mode v2 and missing GA4 conversions. Where those pages give no fact, the check is worded as an action to carry out, not as a statistic.

What does the GA4 setup service add on top of this?

Our GA4 setup page describes a written audit of the property, with read access to GA4, Tag Manager and Ads, followed by fixes in order of impact. The page lists the audit as free with a written deliverable. A self-check cannot do that: it relies on your answers, not on the property itself.

Can the checker tell me whether my GA4 numbers are correct?

No. It shows where your own answers point to a gap. Only a test conversion you watch arrive, and a comparison of GA4 with your CRM, call log or orders, show whether the numbers match what happened. The checker asks whether you have done both.