- Start with one number: compare GA4 leads with your inbox or CRM for the same period.
- Then check tags, events, key events, traffic sources, internal traffic, consent, personal data and settings.
- Fix counting problems before you trust any channel report.
- Data filters are permanent once active, so test them first.
- Note fixes with a date, because old data does not change and comparisons across that date will mislead.
The short answer
To check whether your GA4 data is right, compare one number with a source you trust: GA4 leads against leads in your inbox or CRM, for the same period. Then check that tags load once, that events fire once, that key events are real outcomes, that sources are labelled properly, and that no personal data is getting in. A first pass needs only Admin access and Tag Assistant.
This post is the check for a setup that already exists. To build one, use our tracking setup guide.
Check 1: Compare one number
Pick last month. Count the real leads in your inbox or CRM, leaving out spam and tests. Then look at the same period for your lead event in GA4.
| What you see | Likely cause | Look at |
|---|---|---|
| GA4 a bit lower | Visitors not measured, through consent or blockers | Often normal, so watch the trend |
| GA4 far lower | Event missing on some forms, or tag blocked | Check 3 and check 7 |
| GA4 higher | Double firing, or a click counted as a lead | Check 2 and check 3 |
| GA4 shows none | Broken or never built | Build it first, then come back |
There is no universal right gap, so do not chase a figure from someone else's site. Your own gap should be explainable and stable from month to month.
Check 2: Is the tag on every page, once?
- Open Tag Assistant or Tag Manager preview and load the home page, a blog post, a form page and a thank-you page.
- Look at which tags load. You want one route, usually Tag Manager, and one GA4 measurement ID.
- Check that one page load gives one
page_view. - View the page source and search for
G-andGTM-. An old pasted snippet often sits in a site-wide custom code field.
Two measurement IDs may be on purpose, for example when a parent company has its own property. Find out before you remove one.
Check 3: Do events mean what they say?
Open the Events list in Admin, under Data display. For each event, ask what really fires it.
- Near duplicates. Names like
lead,Lead_Formandgenerate_leadusually mean several people built this over time. - Leads from clicks. Press the form button with an empty form. If the lead event fires, it is counting clicks.
- Calls from clicks. Open your booking link and do not book. If a "booked" event fires, it is not a booking. See tracking booked calls.
- Forms missing out. Send every form on the site, including the small ones in the footer and popups.
For the right way to fire a lead, see tracking website forms step by step.
Check 4: Are key events real outcomes?
Google allows up to 30 key events on a standard property, but a marketing site rarely needs more than two or three. Read the list out loud. If an item is not something your business would pay for, such as a scroll, a page view or a button click, remove the star.
Reports will read differently once a star is gone, so tell whoever reads them, or they will think results have dropped.
Check 5: Are your sources labelled well?
Open the traffic acquisition report and set it to the session default channel group.
- Unassigned. Google's channel rules use this when no rule matches. Look at the source and medium behind it, then fix the links.
- Messy tags. Google's UTM guide says values are case sensitive, so "Meta" and "meta" count as two. Missing values show as (not set). Agree on one lowercase UTM naming list.
- Tools showing as referrals. If a payment page or a tool on another domain appears as a top referrer, add it under unwanted referrals, in the data stream's tag settings.
- Your own domains. If you run more than one, set up cross-domain measurement and test that
_glappears in the URL when you move across.
Check 6: Is your team in the data?
Internal visits and staging tests can inflate small numbers.
- Load tags on the live domain only, so staging stays out.
- In the stream's tag settings, define internal traffic for your office or VPN address.
- Create the data filter in the Testing state first.
- Check the result with the Test data filter name dimension in an exploration. Google says this can take 24 to 36 hours.
- Only then set it to Active.
Be careful with that last step. Google says the effect of an active filter is permanent, and excluded data is never processed. If you need the data later, it is gone.
Check 7: Consent and personal data
Open a private window and reject the banner. Use the Network tab in your browser to see what still reaches Google, then do the same before you answer at all. It should match what you chose: for most setups, nothing or cookieless measurement only. A script pasted outside Tag Manager is the usual cause of a surprise.
Consent law differs by country, and this is not legal advice. The check here is only whether the site does what you decided. Whether the decision is right is a question for someone qualified.
Personal data is the other half. Google's policy does not allow data it could recognise as personally identifiable. It warns that this often gets in by accident through URLs and titles.
- In the pages report, search page paths and query strings for
@andemail=. - Check your search box. Visitors sometimes type emails into it.
- Check thank-you pages for names passed in the URL by a form or booking tool.
- Check event values such as form field text and link text.
If you find some, fix the source first. Google mentions data redaction in Admin as a best-effort help for web streams.
Check 8: Settings people forget
- Data retention. In Admin, under Data retention, standard properties can pick 2 or 14 months for event data. Google says this affects explorations and funnel reports, not standard aggregated reports. Check what yours is set to, because it decides how far back an exploration can go.
- Search Console link. Check that it exists under Product links. Our setup guide covers it.
- Access. Remove past staff and old suppliers from Admin.
- Ownership. Make sure your company owns the account, and not a freelancer's login.
The scorecard and fix order
| Check | Fails when | Fix order |
|---|---|---|
| Personal data | Emails or names in reports | First |
| Consent | Tags ignore your choice | First |
| Tag on every page, once | Double page views or gaps | Second |
| Events mean what they say | Clicks counted as leads | Second |
| Key events | Steps marked as outcomes | Third |
| Sources | Unassigned or messy tags | Third |
| Internal traffic | Team visits in small reports | Fourth |
| Settings | Short retention, old access | Fourth |
Write down the date of each fix in the shared notes. Old data stays as it was, so a comparison across that date can look like a rise or a drop that never happened.
What to do next
Do check 1 today. It takes minutes and it tells you how far to go. If you want a second pair of eyes, our analytics and tracking work starts from $500, or you can send us your site and tell us which report you trust least.
Terms used: GA4 (Google Analytics 4), Core Web Vitals



