A GA4 audit for an inherited client property checks three things in order: whether the property settings are right (time zone, retention, filters, referrals, cross-domain, consent, links), whether the key events fire once on the real action, and whether the reports built on top can be trusted. Work through the checklist below, record what you find in a single sheet, and end with a fix list sorted by how much each problem distorts the numbers the client and the ad platforms rely on.
This guide is written for agency owners, PPC managers and analysts who pick up a new client and find a GA4 property someone else built. It assumes you have Editor access to the property and at least read access to the Google Tag Manager container and the Google Ads account. If you do not have those yet, get them before you start; half of a GA4 audit happens outside GA4.
Before you open GA4: collect the context
An audit without context turns into a list of opinions. Spend fifteen minutes collecting the facts that tell you what the property is supposed to do.
- What the client counts as a lead or a sale. A form submission, a booked call, a phone call, a purchase, a quote request. Write down each one and the page or tool where it happens.
- Every domain and tool in the journey. Main site, landing page builder, booking tool, payment page, checkout, help center, any embedded forms from HubSpot, Typeform, Jotform or Calendly.
- How tags are deployed. GTM, hardcoded gtag.js, a CMS plugin, a Shopify app, or several at once. Several at once is the most common source of duplicate events.
- Where the numbers go. Google Ads imports, Looker Studio reports, monthly client decks. Anything downstream inherits the faults you find.
- The CRM or order system. This is your source of truth when you need to check whether GA4 counts are plausible.
Section 1: Property and data stream settings
Start with the settings that affect every report. They are quick to check and some of them cannot be fixed retroactively, so the earlier you catch them the better.
Property settings
- Reporting time zone. Should match the client's business time zone and, ideally, the Google Ads account time zone. A mismatch shifts conversions across day boundaries and makes daily comparisons look wrong.
- Currency. Should match the currency the client reports revenue in. GA4 converts other currencies sent in the
currency parameter, but only if the parameter is present and valid.
- Industry category and business details. Low impact, but worth correcting while you are there.
Data retention
In data collection and retention settings, event data retention defaults to two months on standard properties. That limit applies to explorations and other non-aggregated analysis, not to standard reports, but it catches agencies out when they try to build a year-over-year exploration. Unless the client has a reason not to, set it to the maximum available (14 months on a standard property). The change only applies going forward.
Data stream and tag settings
- Number of web streams. One web stream per site is normal. Multiple web streams for the same site usually mean someone created a second one during a rebuild and both are collecting.
- Measurement ID in use. Confirm the ID on the live site matches the stream you are auditing. View the page source or use Tag Assistant. Old IDs from abandoned properties are common.
- Session timeout and engaged session timer. Defaults are 30 minutes and 10 seconds. Changes are not wrong, but they should be deliberate and documented.
Section 2: Enhanced measurement
Enhanced measurement collects page views, scrolls, outbound clicks, site search, video engagement, file downloads and form interactions without any extra tagging. Most of it is useful. The form interactions option is the one to look at closely.
With form interactions on, GA4 sends form_start and form_submit automatically. In practice these events fire on newsletter boxes, search forms and login forms as well as lead forms, and form_submit can fire on a submit attempt that fails validation, while missing AJAX forms that never trigger a standard submit. If someone has marked form_submit as a key event, the lead count is unreliable.
Our usual recommendation: turn form interactions off once a proper generate_lead event is in place, or leave it on for engagement analysis but never use it as a key event. Also check that site search is configured with the actual query parameter the site uses (for example s on WordPress), otherwise the search report stays empty.
Section 3: Internal traffic and developer traffic
Internal traffic filtering is a two-part setup, and the second part is the one people forget.
- In the data stream's tag settings, define internal traffic rules by IP address. These rules tag matching hits with
traffic_type=internal.
- In data filters, the internal traffic filter must be set to Active. A filter in Testing state only labels data; it does not exclude it.
Also check the developer traffic filter, which excludes hits sent with debug mode (from GTM preview or Tag Assistant). If it is inactive, every test conversion your team or the previous agency ran is sitting in the reports.
Two cautions. Active filters permanently exclude data; there is no way to get it back. And IP filtering does little for remote teams on home connections, so do not promise the client that internal traffic is fully excluded.
Section 4: Key events inventory
This is the core of the audit. Export or list every event in the property (the Events report under admin or a free-form exploration of event name by event count works), then mark each key event.
| Column | What to record |
| Event name | Exact name, including case |
| Source | Automatic, enhanced measurement, GTM tag, gtag in code, created in GA4 UI, plugin |
| Trigger | What action actually fires it |
| Key event? | Yes or no, and whether it should be |
| Count, last 30 days | Compare with CRM or order counts |
| Parameters | Value, currency, form or lead type, transaction ID |
| Verdict | Keep, rename, fix trigger, remove key event flag, delete tag |
Typical findings in an inherited property:
- Page views of a thank-you page marked as key events, which count every refresh.
- Events created in the GA4 interface with "create event" rules that copy
page_view under a new name, then marked as key events.
- Several names for the same thing:
lead, Lead, form_submit, generate-lead. GA4 event names are case-sensitive, so these are different events.
- Key events that have not fired in months because the form changed.
- Too many key events. A property has a limit on key events, but the real problem comes long before the limit: when everything is a key event, conversion rate means nothing.
Our naming and selection approach is covered in detail in our GA4 key events and naming plan.
Section 5: Duplicate events
Duplicate events inflate key event counts and, if those key events are imported into Google Ads, feed bad data into bidding. Check for them in three ways.
- Count comparison. If GA4 reports noticeably more leads than the CRM received over the same period, look for duplication before anything else.
- Tag inventory. Look for the measurement ID in GTM, in the site's code, and in CMS or e-commerce plugins. Two Google tags with the same ID will send every event twice.
- DebugView. Submit a test lead and watch for the event appearing twice within the same second, or once from GTM and once from a plugin with slightly different parameters.
For purchases, check whether transaction_id is present. GA4 uses it to deduplicate purchases with the same ID, so a missing transaction ID turns every order confirmation reload into another sale.
Section 6: Unwanted referrals and cross-domain
Open the traffic acquisition report and filter for session source. If you see the client's own domain, a payment provider (Stripe, PayPal), a booking tool (Calendly) or a checkout domain as a referral source, sessions are being broken and conversions are being credited to the wrong source.
- Domains the user passes through and returns from (payment pages, booking tools) go in the unwanted referrals list in the stream's tag settings.
- Domains the client owns where the journey continues (a separate landing page domain, an app domain) go in configure your domains, which enables the cross-domain linker.
This is a large enough topic that it has its own guide: GA4 cross-domain tracking and self-referrals.
Section 7: Consent mode
If the client has EEA or UK traffic, consent mode is not optional for anyone who wants Google Ads measurement to keep working properly. Check:
- Is a consent banner present, and is it connected to Google's consent mode at all?
- Are all four parameters set:
ad_storage, analytics_storage, ad_user_data, ad_personalization?
- Is a default state set before any Google tag fires, and does the banner send an update when the user chooses?
- Is it basic or advanced mode, and does the client understand the trade-off?
Tag Assistant's consent view shows the default and update states for each page load. A banner that looks fine but never updates consent is a common fault. The full setup is in our Consent Mode v2 implementation guide. Consent is a legal question as well as a technical one; the client's legal advisor decides what the banner must do.
Section 8: Google Ads and Search Console links
- Google Ads link. Confirm the correct Ads account is linked (agencies often find a link to a previous agency's MCC or to a closed account). Check that auto-tagging is on in Google Ads, otherwise GA4 sees paid clicks as organic or unattributed.
- Imported key events. In Google Ads, check which conversion actions come from GA4 and whether they are primary. A GA4 import and a native Google Ads tag counting the same lead, both set as primary, double the conversions bidding sees.
- Search Console link. Linked to the right web stream, so organic query data is available.
- Google signals. Note whether it is on. It affects cross-device reporting and can trigger data thresholds in some reports.
For the Google Ads side of the audit, our Google Ads conversion tracking setup guide covers counting, windows and primary versus secondary actions.
Section 9: Attribution settings
In attribution settings, record the reporting attribution model and the lookback windows. The options available as of 2026 are data-driven, paid and organic last click, and Google paid channels last click. The older rule-based models such as first click and linear were removed. Acquisition key event windows are 7 or 30 days; other key events can be 30, 60 or 90 days.
None of these is wrong in itself, but the client should know which one is in use, because it changes how much credit each channel gets in reports. If the model was changed recently, note the date: historical reports will not line up with old screenshots.
Also check the reporting identity setting. Blended is the usual choice and is required for consent mode behavioral modeling to appear in reports.
Section 10: Naming, custom dimensions and reporting hygiene
- Event and parameter names in one consistent style (lowercase with underscores matches GA4's own recommended events).
- Custom dimensions registered for the parameters people actually report on, such as
form_name or lead_type. Parameters that are not registered as custom dimensions do not appear in standard reports.
- No personal data in event parameters or page URLs. Email addresses in query strings after a form submit are a common leak and a policy problem.
- Channel groups: check whether a custom channel group exists and whether paid social, email and affiliate traffic are classified correctly, which depends on consistent UTMs. Our UTM naming convention for agencies covers that side.
Section 11: DebugView and live checks
Settings tell you what should happen. Testing tells you what does. For every conversion path you listed at the start:
- Open GTM preview (or enable debug mode another way) and complete the action as a real user would, on desktop and on a phone.
- In DebugView, confirm the key event appears once, with the expected parameters: value, currency, form name, transaction ID where relevant.
- Confirm the page location does not contain personal data.
- Repeat with consent denied (if a banner is in place) and confirm tags behave as the consent setup intends.
- Check the realtime report as a second view, then confirm in standard reports the next day.
Save a screenshot of each test. A DebugView log is the evidence that a fix works, and it is what the client or your account manager will ask for when numbers move.
Turning findings into a prioritized fix list
An audit is only useful if it ends with a short list of what to do first. We sort every finding into three tiers.
| Priority | What belongs here | Examples |
| 1. Fix now | Anything that feeds wrong data to bidding or to the client's lead count | Duplicate key events, page view key events imported to Ads, wrong Ads account linked, missing transaction IDs, self-referrals stealing paid credit |
| 2. Fix this month | Gaps that lose data or distort reports but do not mislead bidding directly | Consent mode defaults missing, AJAX forms not tracked, internal traffic filter in testing state, data retention at two months |
| 3. Clean up | Hygiene that makes reports readable | Naming inconsistencies, unregistered custom dimensions, unused events, channel grouping |
For each item, write the finding, the evidence (a screenshot or count), the fix, who does it (agency, client developer, or tracking partner) and how it will be verified. Keep the language plain enough that the client's marketing manager can read it.
Finally, record the date each fix goes live. GA4 does not rewrite history, so the client needs to know that leads before a given date were counted one way and after it another. That change log is what stops a month-over-month report from turning into an argument.
Common mistakes during GA4 audits
- Auditing only GA4. Many faults live in GTM, the CMS, or Google Ads. Look at all of them.
- Deleting events before replacing them. Remove the key event flag from a bad event only once the correct event is live and tested, otherwise the Ads import goes to zero.
- Trusting event counts without a CRM comparison. A count that looks reasonable can still be half duplicates and half missing.
- Changing attribution settings mid-month without telling anyone.
Need a second pair of eyes on a client property?
If you would rather hand the property over than work through the checklist yourself, our GA4 audit covers every section above, white-label, starting with a free tracking check and ending in a prioritized fix list with DebugView evidence.