A GA4 and Google Ads discrepancy does not always mean a broken tag. Reports may use different events, channels, dates or counting rules. Compare the same customer action and reporting period first. Once those definitions match, investigate tags and duplicate records for the remaining difference.
Compare the same action first
Choose a specific outcome such as a successful quote request, purchase or appointment. Record the event name, conversion action, accounts, report and date range. Avoid comparing page views with submissions or all-channel outcomes with advertising-related outcomes.
Imagine a hypothetical website reporting enquiries from every channel while an Ads report displays one advertising conversion action. Different totals are unsurprising, but the settings and data still need to confirm the explanation. The first task is to document scope rather than immediately changing tags. A reproducible comparison gives the technical team a much clearer starting point.
Separate key events from ad conversions
Google Analytics describes a key event as an event measuring an important business action, while a Google Ads conversion supports advertising performance measurement and optimisation. A standard GA4 key-event total should therefore not automatically be treated as equivalent to a selected Ads conversion column.
Create a measurement dictionary for the action. Clicking a submit button, successfully submitting a form and becoming a sales-qualified enquiry are different stages. Give each a purpose. If the same outcome is collected through two methods, document how both measurements should be used in reporting and optimisation rather than assuming both belong in every total.
Match dates, channels and counting settings
Illustrative example: a customer clicks an ad on 10 October and submits a form on 12 October. A standard Ads conversion column can associate the result with the ad interaction date, while a conversion-time column uses the day the action occurred. Comparing only 12 October can therefore produce different totals. Check the definition of the actual column in your account.
Then record the time zones, included conversion actions and channel scope. Comparing all-channel GA4 key events with Ads-attributed results introduces a definition difference. Do not treat the newest reporting period as final while its data is still processing.
| Check | Write down side by side | Possible source of the difference |
|---|---|---|
| Action | GA4 event name and Ads conversion action | A click and a completed form are different actions. |
| Date | Standard versus conversion-time column | Ad interaction and completion may fall on different days. |
| Channel | All traffic versus ad-attributed results | The reports may cover different sets of results. |
| Counting | Every action versus one-action rule | Repeat actions can be counted differently. |
| Freshness | When the report was retrieved | Recent results may still be processing. |
Investigate tags for the remaining gap
Once definitions align, follow a controlled action through the journey. Did the site complete the task, create the business record, trigger the expected event and associate it with the intended account? Different forms, subdomains or payment returns may behave differently. Record the conditions for each path.
Keep personal customer information out of general debugging sheets. The action type, time and an appropriately limited record identifier can be more useful for technical discussion. Explaining the reproducible path and conditions is clearer than circulating large collections of customer records.
Explain what each number represents
Show the compared metrics, the explained difference and the remaining uncertainty. Operational records may answer how many enquiries arrived, while advertising reports answer a question about attributed contribution. Make that distinction explicit rather than expecting one metric to represent the entire journey.
Repeat the same check after a correction. Record previous and updated settings, the review period and the decision owner. Known differences then stop being rediscovered as new errors every month, while a genuinely new measurement issue becomes easier to identify.