If Google Ads and GA4 show different conversion totals, first check whether you are comparing the same action, reporting column and date basis. A difference can come from reporting settings, delayed data or a genuine tracking fault. Changing the campaign budget before separating these possibilities can lead to the wrong decision.
The useful question is not “Which platform is right?” It is “Can we explain what each number includes, and verify that the intended business action is recorded correctly?” Here is the order I would use for that review.
Start with the same action and report
Write down the exact Google Ads conversion action, its source and the corresponding GA4 event. “Lead” is not precise enough: a button click and a successfully received enquiry can have similar names but mean different things.
GA4 distinguishes key events from conversions shared with Google Ads. Standard Reports and Explore let you analyse key events; advertising conversion reporting serves a different comparison. Do not treat these labels as interchangeable. Google's explanation of conversions and key events.
There is also a current option worth checking: GA4's Conversion performance report, still in beta and gradually rolling out. If available, its Google Ads perspective uses the selected linked account's settings. Its All conversions count attributed to that account is comparable with All conversions in Google Ads. The report also lets you choose interaction-time or conversion-time attribution. Check this before assuming every Ads–Analytics difference is unavoidable. Conversion performance report documentation.
Match the account, action and filters. Keep the channel scope identical, too.
Align the date basis before comparing a month
Google Ads' standard conversion columns assign results to the ad interaction date. It also offers columns such as Conversions (by conv. time) and All conv. (by conv. time), which use the conversion date. Choose the column that matches the other report's basis, and check the account or property time zones. Understanding Google Ads conversion data.
An illustrative example: someone clicks an ad on 30 September and orders on 2 October. A September report organised by interaction date can include that conversion; a report organised by conversion date places it in October. Neither result, on its own, proves a missing order.
Save the chosen date range and timing setting with the export. A screenshot headed “September conversions” is not enough to reproduce a comparison.
Separate counting rules from attribution rules
Counting asks how many occurrences count. Attribution asks which interactions receive credit. Investigate them separately.
In Google Ads, One and Every apply to the conversion action after an ad interaction. “One” does not combine several different actions into a single lead. In GA4, key-event counting can be once per event or once per session. These are different rules, even when their names sound similar. Google Ads counting options, GA4 key-event counting.
Then record the attribution model, conversion window and channels eligible to receive credit. For Analytics-based web conversions, Google documents a choice between Google paid channels and paid and organic channels. Changing it can affect reporting and bidding in linked Ads accounts. It is not a cosmetic adjustment to make two totals agree. Channels receiving conversion credit.
Document the existing settings before changing anything. If the real question is which outcome should influence bidding, use the separate guide to choosing Google Ads lead-generation conversions.
Allow processing time, then investigate missing signals
Google advises allowing 24–48 hours for data processing and synchronisation before comparing Ads and Analytics. That is not a promise that a recent period is commercially complete: a customer can still enquire or buy later after an earlier interaction. Record when you exported each report. Google's discrepancy troubleshooting guide.
Also read any data-quality or modeling indicators. GA4 behavioral modeling has eligibility requirements; not every property receives it. Estimates are not evidence that every unobserved action has been recovered. Google's behavioral-modeling documentation.
There is no single percentage gap that proves your setup is healthy. A small difference can conceal a broken path used by a small audience; a large difference may have an identifiable reporting explanation. The job is to explain the difference, not to find a convenient tolerance.
Trace the measurement against a business record
Use controlled tests to follow the intended action from the website to its operational record and tracking event. Google recommends Tag Assistant for checking the relevant tags in its measurement troubleshooting guidance. You do not need to click your own live ads just to test whether an event fires.
Include these scenarios in the test brief:
- Successful action: a received enquiry or accepted test order triggers the intended event at the agreed stage.
- Failed action: a rejected form submission or unsuccessful checkout does not masquerade as a completed result.
- Repeat visit: reloading the confirmation page or returning to it does not create an unintended duplicate.
- Changed path: an external booking step or payment redirect does not silently interrupt the measurement you intended to preserve.
Use an approved test or staging flow and distinguish test records from genuine business activity. Keep customer details out of shared screenshots.
The record depends on the product. My Quattro Hunting work includes an online shop and order management; SpainProperty includes property browsing and consultation booking. These are different commercial journeys. They illustrate why an audit must first define the relevant business stage; they are not evidence of a particular tracking setup or campaign result.
An order system confirms that an order exists. It does not, by itself, prove which advertising interaction deserves credit. Likewise, a test event appearing correctly does not prove that an entire month's attribution is correct.
Use a comparison worksheet someone else can reproduce
Ask for one short record per action, containing:
- Identity: Ads account and action, conversion source, GA4 property and event name.
- Report scope: exact report, column, filters and selected account or attribution perspective.
- Time: date range, interaction or conversion date, time zones and export timestamp.
- Rules: counting method, attribution model, window and eligible channels.
- Evidence: relevant setting screenshots, test outcomes and the matching operational stage.
- Conclusion: explained reporting difference, demonstrated implementation fault, or unresolved question with a next check.
This makes the review portable between a campaign manager and a developer. “Tracking looks fine” is not a useful handover. “The repeated confirmation-page visit emits another purchase event; here is the test and the proposed correction” is something a developer can act on.
Commission a focused review before a wider rebuild
A sensible first scope is one important conversion journey, a reproducible comparison and a documented test plan. Ask for a list of findings, the evidence behind each one, any proposed fixes and a record of what changes. Recheck the affected journey after a fix before interpreting a rise or fall as a business result.
If you want help connecting the website, advertising and enquiry process, send me your website and the two reports you are comparing. You can reach me at hello@davidbuka.com through the contact page. Start with the report names and intended action; account credentials and customer messages are not needed for the initial enquiry.
Platform documentation checked on 26 October 2026. The date example is illustrative. The cover is an original editorial illustration, not a photograph of client records.