Conversion Tracking Audit: Why Your Marketing Numbers Don’t Match

Here is the moment this article is about. Someone asks a simple question in a meeting: how many leads did the campaign bring in last month. GA4 says 312. Google Ads says 487. The CRM says 240 that a salesperson actually called. Three systems, three numbers, one campaign, and a budget decision waiting on an answer nobody in the room can give with a straight face.

The reflex is to treat this as a bug. Someone is told to find the error and make the numbers agree. They will not agree, because they were never built to. GA4 and the ad platforms answer different questions with different methods, so chasing a perfect match is hunting for a configuration error that does not exist, while the real faults, the ones that genuinely corrupt the data, sit undisturbed.

A conversion tracking audit separates the two. It tells you which part of the disagreement is methodology you should expect and leave alone, and which part is a break you need to fix. Then it reconciles all three systems to one number you can put in a board deck and defend. The rest of this piece is how that works.

Where tracking breaks

Some of the gap between your systems is design, covered in the next section. This section is the other kind: the genuine faults, the places where a conversion is lost, duplicated or attributed to the wrong thing because something is misconfigured. These are what an audit hunts for first, because they corrupt the data rather than just reframe it.

Tags are the first place to look. A conversion tag that fires twice counts one sale as two. A tag that fires on a page load rather than the actual conversion event counts intent as completion. A tag missing from one step of a checkout or form flow loses every conversion that passes through it. These are not subtle once you go looking, but they are invisible from the dashboard, because a dashboard reports the number it was given without questioning whether the number is real. Tags patched in over years, as forms and booking tools and checkout flows changed, are where most of this lives.

Consent is the second, and in 2026 it is part of the measurement system rather than a banner bolted on top. When a user rejects tracking, Consent Mode changes what gets collected, and in regions with high rejection rates GA4 can undercount by 20 to 40 percent against the ad platforms, which model the denied-consent conversions from a larger signal pool. A consent setup that is misconfigured, and most are, either loses data it was allowed to keep or collects data it was not, and both corrupt the numbers in ways that look like a tracking fault because the effect is the same: the count is wrong.

Cross-device and the state of the cookie are the third. The third-party cookie did not die on a date, whatever the headlines said. Google abandoned forced deprecation in 2025 and moved to user choice, so the cookies still exist in Chrome, but Safari and Firefox have blocked them for years, and roughly a third of traffic was already cookieless before you add ad blockers on top. The practical effect has been true for a long time: a user who clicks an ad on their phone and converts on their laptop is two anonymous sessions to a cookie-based system, not one journey. The cookie did not break your cross-device tracking this year. It has been lossy for years, and the only question is whether your setup accounts for that or pretends otherwise.

Double-counting is the fourth, and it is the one teams create themselves. Run a native Google Ads conversion tag and import the same conversions from GA4, and the same sale can be counted by both. Add a Meta pixel and an email tool that each claim the conversion, and one lead becomes three across four dashboards. Every platform credits itself, so a conversion touched by three channels is counted three times if nothing deduplicates it. This is why the totals across your platforms can exceed the number of sales you actually made, which is the clearest possible sign that the stack is measuring its own reflections.

Reconciling platform, analytics and CRM

The faults above are fixable. The disagreement that remains after you fix them is not a fault, and trying to eliminate it is the mistake. This section is about understanding it and reconciling around it.

Three structural differences produce most of the gap between GA4 and an ad platform, and all three are working as designed. The first is the counting method: an ad platform can count every conversion after a click, so three purchases from one user count as three, while GA4 counts events its own way. The second is timing: Google Ads attributes a conversion to the date of the click, GA4 to the date the conversion happened, so a Monday click and a Friday purchase land in different periods.

The third is the attribution model. An ad platform gives full credit to its own click, while GA4’s data-driven model splits credit across every channel that touched the journey. A sale the ad platform counts as one might be three-tenths of a conversion in GA4, because organic and email also played a part.

None of those is wrong. They are three honest answers to three different questions. The ad platform is answering how many conversions its ads drove. GA4 is answering how many conversions happened and what contributed to them. Asked different questions, they give different numbers, and forcing them to match destroys the information each was built to give you.

So you do not reconcile by making them equal. You reconcile by anchoring to the one system that measures the thing you actually care about: money. The CRM is the only system that knows which leads became opportunities and which opportunities became revenue. An ad platform knows clicks. GA4 knows site events. Only the CRM knows what was real, because a salesperson worked it and it either closed or did not.

This is the move most teams skip. They compare GA4 to the ad platform, see the gap, and argue about which marketing tool to believe, when both are marketing tools measuring the top of the funnel. Comparing two of them to each other validates nothing. Comparing both to the CRM, where you can see how many of those tracked conversions became actual opportunities, tells you what your tracking is worth. The reconciliation is not GA4 against Google Ads. It is both of them against closed revenue, with the expected methodology gaps acknowledged rather than fought.

Enhanced and offline conversions

Two mechanisms close specific parts of the gap, and it is worth being precise about what they fix, because they are often sold as a cure for everything and they are not.

Enhanced conversions address the signal lost to blocked cookies and rejected consent. Instead of relying on a browser cookie that may never be set, they send hashed first-party data, an email or phone number the user already gave you, from your own systems to the ad platform, which matches it to a logged-in account to confirm the conversion. This recovers conversions that browser-based tracking simply never sees, which is why it has moved from optional to standard. It is a partial answer to the cookie erosion described above, not a replacement for clean tags.

Offline conversions address a different gap: the one between a form fill and a real customer. In any business with a sales cycle, the conversion that matters happens after the website, in the CRM, when a lead qualifies or a deal closes. Offline conversion tracking sends that later event back to the ad platform against the original click, so the bidding learns from qualified outcomes rather than raw submissions. It is the mechanism that connects the CRM anchor back to the platform that needs to optimise toward it, which is why it sits at the centre of a measurement setup rather than at the edge.

Both share a foundation, and it is the same foundation that makes the whole stack trustworthy: server-side tracking. Collecting conversion data on your own server first, then sending it to each platform through their APIs, means the data is captured before a browser or an ad blocker can drop it, and every platform receives the same source data. Get there and your remaining discrepancies are pure methodology, the expected gaps, rather than a mix of methodology and data loss you cannot tell apart. That is the difference between a stack you reconcile and a stack you guess at.

A measurement setup you can trust

A setup you can trust is not one where every number matches. It is one where you know exactly why each number says what it says, and you have one anchor you measure everything against.

In practice that means four things hold at once. The tags are clean, firing once, on the real conversion event, on every step of every flow. The consent configuration is correct, keeping what it is allowed to keep and losing nothing it is not. Deduplication is deliberate, so one conversion is counted once no matter how many platforms want to claim it. And the CRM is wired back into the platforms through offline conversions, so the system optimises toward revenue rather than toward form fills.

When those hold, GA4 still reports its number, the ad platform still reports its higher one, and the CRM still reports the smallest and truest of the three. The difference is that you can now explain every gap between them, and you have one figure, closed revenue, that every decision rests on. The disagreement stops being a problem to solve and becomes information you read.

When systems disagree, the decision rests on partial truth

This is a measurement problem, but underneath it is a principle that applies anywhere systems are supposed to describe the same reality and do not. When two systems disagree, every decision made on either one rests on partial truth, and the answer is never to pick the number you like. It is to understand what each system actually measures, find the single source that maps to the outcome you care about, and reconcile the rest to it.

In marketing that source is closed revenue and the system is the CRM. In a regulated business it might be a different anchor entirely, but the discipline is the same: disagreement between systems is not noise to be averaged away, it is a signal that you have not yet found the number that maps to truth.

FAQs

Why do my marketing numbers not match?

Because GA4, the ad platforms and the CRM measure different things in different ways. Ad platforms count conversions their ads drove, attributed to the click date, with full credit to their own click. GA4 counts site events on the conversion date and splits credit across channels. The CRM counts what became a real opportunity or sale. A 20 to 30 percent gap between the analytics and the ad platform is normal and expected, not a fault. The numbers not matching is usually the systems working as designed, with a smaller portion caused by genuine breaks like double-counting or broken tags.

What is a conversion tracking audit?

A structured review of how conversions are captured, counted and attributed across your analytics, ad platforms and CRM. It separates the discrepancies that are expected methodology from the ones that are genuine faults, such as tags firing twice, conversions lost to consent misconfiguration, cross-device journeys broken by cookie loss, or the same sale counted by several platforms. The output is a measurement setup where you understand every gap and can reconcile all systems to a single source of truth.

Why do GA4 and ad platforms disagree?

Three structural reasons, all by design. Counting method: an ad platform can count every conversion after a click while GA4 counts events differently. Timing: ad platforms attribute to the click date, GA4 to the conversion date, so the same sale lands in different periods. Attribution model: an ad platform gives full credit to its own click while GA4 splits credit across every channel in the journey. Consent and modeling widen the gap further, since ad platforms model denied-consent conversions from a larger signal pool than GA4 has. None of these is an error.

How do I reconcile platform and CRM data?

Stop comparing the marketing tools to each other and anchor both to the CRM. The ad platform and GA4 both measure the top of the funnel, so comparing them validates nothing. The CRM is the only system that knows which tracked conversions became real opportunities and revenue. Reconcile by checking how many of the conversions each platform reports actually appear as opportunities in the CRM, accept the methodology gaps between the platforms as expected, and use closed revenue as the single figure every decision rests on. Offline conversion tracking then feeds that CRM outcome back to the platforms so they optimise toward it.

What breaks conversion tracking?

Four things, beyond the expected methodology differences. Tags that fire twice, fire on the wrong event, or are missing from a step in the flow. Consent misconfiguration that loses data it was allowed to keep or collects data it was not. Cross-device and cookie loss, since Safari and Firefox block third-party cookies and a large share of traffic is already cookieless, so a phone-to-laptop journey reads as two anonymous sessions. And double-counting, where a native ad tag plus a GA4 import plus other platform pixels each claim the same sale. The first and last are the most common and the most fixable.

How do I trust my numbers?

Trust does not come from the numbers matching, because they will not. It comes from four things holding at once: clean tags firing once on the real event, correct consent configuration, deliberate deduplication so one conversion is counted once, and the CRM wired back to the platforms through offline conversions. Server-side tracking underpins all of it by capturing data before a browser can block it. When those hold, you can explain every gap between your systems and you have one anchor, closed revenue, that every decision rests on. That is a setup you can trust, even though the numbers still differ.

Last reviewed: June 2026

This article provides general information about conversion tracking and marketing measurement. Platform behaviour, attribution models, consent rules and privacy regulations change over time. Verify current platform documentation and your own regulatory obligations before making measurement decisions.