It is one of the most common questions in any analytics audit: Google Ads shows one number of conversions, GA4 shows another, and neither matches the CRM. In most cases none of the systems is broken. They are measuring the same activity with different rules.
1. Different attribution models
Google Ads credits conversions to Google Ads clicks. GA4 distributes credit across every channel a user interacted with, using data-driven attribution by default. A purchase that followed a Google Ads click and later an email click can be fully credited in Google Ads and only partly credited to paid search in GA4.
2. Conversion date versus click date
Google Ads reports a conversion against the date of the ad click. GA4 reports it on the date the conversion happened. A click on Monday that converts on Friday appears on Monday in Google Ads and on Friday in GA4, so short date ranges rarely line up.
3. Counting settings
A Google Ads conversion action can count every conversion or one per click. Lead forms usually use one, purchases usually use every. GA4 key events count each time the event fires. If the settings differ, the totals will too.
4. Conversion windows and view-through
Each Google Ads conversion action has its own click-through window, commonly 30 days, and some campaign types also record engaged-view or view-through conversions. GA4 has its own lookback window and does not include view-through conversions from Google Ads.
5. Tag source and consent
If Google Ads uses its own conversion tag while GA4 uses a separate event, a trigger that fires in one but not the other creates a real gap. Consent Mode also affects both, and Google Ads may include modelled conversions where consent was declined.
How to check your setup
- Compare the same conversion over a longer range, such as 30 days, to reduce timing effects.
- Confirm the counting setting and conversion window on each Google Ads conversion action.
- Test the conversion in Tag Manager Preview and confirm both the Google Ads and GA4 tags fire once.
- Use one source of truth for the conversion event, ideally the same data layer event for every platform.
A difference between platforms is expected. A difference you cannot explain is the one worth fixing.