Two systems, two questions
Google Ads reporting exists to answer one question: what did the ads cause, so that bidding and budget decisions can be made? It therefore counts conversions it can join to an ad interaction, applies the AttributionAssigning conversion credit to touchpoints. model and conversion window set on each conversion action, and reports the result against the date of the interaction. It will credit itself with a conversion that GA4 assigns to email, because from its point of view the ad click happened and the conversion followed within the window.
GA4 exists to describe what happened on the site across every channel. It records sessions and events, assigns each session a source and medium from the landing parameters, and attributes conversions across channels under its own model and lookback settings. It cannot see ad impressions or engaged views, does not use Google Ads' cross-device or enhanced matching, and reports conversions on the day they happened. It will assign to organic search a purchase that followed an ad click a fortnight earlier, because the session that converted came from a search result.
Neither is wrong. The measurement hub describes what Google Ads counts; the analytics hub covers GA4 and the wider measurement stack. The practical consequence is that every account needs to know which question it is asking before it picks a number.
Google Ads conversion reporting vs GA4 conversion reporting
| Row label | Google Ads conversion reporting | GA4 conversion reporting |
|---|---|---|
| Traffic unit | Clicks on ads, with invalid clicks filtered and not billed | Sessions, started when the first event fires; one session may follow several clicks, and a click may never start a session |
| Conversion unit | Conversion actions with their own counting rules ("one" or "every") | Key events, counted once per event or once per session depending on the event's counting setting |
| Revenue | Value passed in the conversion tag or import, in the action's currency handling | Purchase revenue from the ecommerce event, in the property's reporting currency |
| Attribution scope | Credit only among Google Ads interactions; other channels are invisible | Credit across all channels; Google Ads is one source among many |
| Model | Data-driven or last click, set per conversion action | Data-driven or last click variants, set per property, with paid-and-organic rules |
| Window | Click-through, engaged-view and view-through windows set per action, measured from the ad interaction | Lookback windows set per property, measured from the conversion backwards |
| View-through and engaged view | Counted in their own columns for display and video | Not available; only sessions and their sources exist |
| Cross-device and matching | Signed-in matching, enhanced conversions and offline imports extend what can be joined | Google signals can link devices where enabled; no enhanced conversion matching |
| Consent and modelling | Consent mode modelling fills gaps in ad-attributed conversions | Behavioural modelling fills gaps in sessions and key events, with its own thresholds |
| Date logic | Conversion reported on the date of the click or interaction; recent days fill in later | Conversion reported on the date of the event; the period is complete when it ends |
| Tagging dependency | Auto-tagging click identifier and the Google tag or conversion linker | GA4 tag and, for channel labels, the auto-tagged identifier or manual UTMURL tags identifying traffic source. parameters |
| Purpose | Running bidding and campaigns | Understanding the site and the full journey |
The causes in detail
Clicks and sessions. A click is recorded when the ad is clicked; a session is recorded when the GA4 tag on the landing page fires. People who click and abandon before the page loads, who block the analytics tag, or who decline consent produce clicks without sessions. People who click twice in quick succession produce two clicks and one session. Redirects that strip the click identifier produce sessions that GA4 labels as something other than Google Ads. Clicks therefore run above sessions on most accounts, and a large gap points at page speed, consent or redirect problems.
Attribution scope. Google Ads only distributes credit among its own interactions, so a conversion with an ad click anywhere in its path within the window is a Google Ads conversion. GA4 distributes credit among every channel in the path. The same purchase can be one full conversion in Google Ads and a fraction of one in GA4's channel report, or zero if GA4's model gives it to a later email click. This is the largest single cause on most accounts and it is structural, not fixable.
Windows. Google Ads windows run forward from the interaction; GA4 lookback windowsThe time after an ad interaction in which a conversion can still be credited to it. run backward from the conversion. The lengths differ by default and are set in different places. A conversion that falls inside one window and outside the other is counted by one system only. How to set the Google Ads side for each conversion action, from the account's own conversion lag, is covered in the guide to conversion windows and attribution settings.
View-throughA conversion credited to an ad that was shown but not clicked, within a set view window. and engaged view. Display and video campaigns receive view-through and engaged-view conversions in Google Ads, in their own columns. GA4 has no concept of an impression, so these conversions appear under whatever session source brought the person to the site. Accounts running video and display see the gap widen in proportion.
Consent and modelling. Both systems model conversions that consent prevented them from observing, but they do so independently, with different inputs and different thresholds for when modelling applies. The modelled components do not reconcile with each other and are not meant to.
Timing. A Google Ads report for last week keeps growing for as long as the conversion window is open, because conversions are placed on the click date. A GA4 report for last week is final once processing completes. Comparing the two for a recent period compares an unfinished number with a finished one.
Tagging faults. Everything above is by design. Faults are different: auto-taggingA Google Ads setting that adds a click identifier (GCLID) to every ad click's landing URL. switched off, the conversion linker missing, duplicate actions, a GA4 key event counted per event while the Google Ads action counts once, or a GA4 import running alongside a native tag. The conversion tracking tutorial covers the Google Ads side of these; faults should be excluded before any structural explanation is accepted.
Reconciliation method
Reconciling is explaining, not matching. Done monthly, it takes an hour once the sheet exists.
- Fix the period. Choose a month that ended long enough ago for the longest Google Ads conversion window to have closed, so both numbers are final.
- Rule out faults. Confirm auto-tagging is on, the conversion linker fires, no action is duplicated, counting settings match between the GA4 event and the Google Ads action, and no import runs alongside a native tag. Any fault found ends the exercise until fixed.
- Align the units. Compare Google Ads clicks with GA4 sessions from the Google Ads source, and Google Ads primary conversions with the equivalent GA4 key events, under matching counting rules.
- Remove what GA4 cannot see. Subtract Google Ads view-through and engaged-view conversions, then cross-device conversionsA journey in which the ad interaction and the conversion happen on different devices, so that only a signed-in identity or modelling can connect them., from the Google Ads figure; these have no GA4 counterpart.
- Set GA4 to the nearest model. Read the GA4 advertising attribution reports with a last-click model limited to Google paid channels, which is the closest GA4 can get to Google Ads' own view.
- Attribute the remaining gap. What is left is the combined effect of window differences, journeys with other channels, consent modelling and click-to-session loss. Estimate each from the available reports and accept that the final residual will not be zero.
- Record it. A line per cause, per month, in a sheet. Causes that are stable are structural; a cause that jumps is a fault or a configuration change, and change history will show which.
A monthly reconciliation for an online course provider
In a closed month, Google Ads reports 1,180 enrolments and GA4's advertising report credits Google Ads with 760 under paid-channel last click. The analyst removes 95 view-through enrolments from YouTube campaigns and 60 cross-device enrolments, bringing Google Ads to 1,025. Comparing a 90-day click window with GA4's shorter default lookback explains roughly another 110, because the courses sell slowly. Journeys that ended on an email click account for a further 90 under GA4's model. The remaining gap of about 65 is put down to consent refusals and clicks that never became sessions, which the click-to-session ratio of around 88 percent supports. Nothing is a fault. The following month the gap widens by 300 and the sheet shows it is entirely in the view-through line, which change history explains: a new display campaign launched.
Which number to use for which decision
| Decision | Use | Because |
|---|---|---|
| Setting and judging bid targets | Google Ads conversions and value | Smart Bidding optimises towards this count; judging it on another is incoherent |
| Shifting budget between Google Ads campaigns | Google Ads | Same scope, same rules across campaigns |
| Shifting budget between channels | GA4 channel reports, with caution | Only GA4 sees all channels, but its model still favours the last session |
| Understanding on-site behaviour after the click | GA4 | Engagement, pages, funnels and conversion rate (CVRShare of visits or clicks that convert.) by landing page exist only here |
| Reporting to finance | Back-office orders or CRM outcomes, with both platforms as supporting detail | Neither platform is the ledger |
| Deciding whether ads caused growth | Neither alone; an IncrementalityConversions caused by marketing. test | Attribution of any kind measures correlation along a path, not causation |
The conversion rate calculator keeps rate comparisons honest when the denominators differ: clicks in one system, sessions in the other. The author's paid media analytics expertise page describes how a reconciliation sheet fits into regular reporting, and the enhanced conversions guide explains one reason the Google Ads figure tends to rise relative to GA4 over time. The same gaps carry through to value: how Google Ads calculates ROAS shows what its conversion value / cost includes and why GA4's figure for the same campaigns is usually lower.
Common mistakes
- Treating the gap as an error. Most of it is structural and will never close.
- Comparing a recent period. The Google Ads number is still growing.
- Comparing clicks with users, or conversions with sessions. Units first.
- Judging bidding on GA4. The algorithm never saw those numbers.
- Switching Google Ads to GA4 imports to make the discrepancy disappear. That adopts one system's limits rather than understanding both.
- Reporting whichever number is higher. Finance will eventually ask for the ledger.
- Ignoring a gap that suddenly changes. Stable gaps are structure; moving gaps are faults or launches.
- Reconciling before the window closes. Pick a period old enough for the longest conversion window to have finished, or the Google Ads side is still moving while you compare.
- Forgetting that the modelled parts never match. Each system fills consent gaps on its own; the two estimates are not supposed to agree and should not be compared line by line.
