Marketing Data Doesn’t Match? Which Number to Believe

None of your tools is lying, and there is no single true number to find. Each system counts a different population of people, on a different clock, under different crediting rules, so disagreement is the expected output of a correct setup. The number you report is the one whose counting rules match the decision you are about to make: the ad platform's number for bidding, your analytics for cross-channel behavior, your own back office for money.

That answer is unsatisfying the first time you hear it, because you came looking for the real figure. There isn't one. What there is instead is a way to pick deliberately, defend the pick in a meeting, and tell the difference between a gap that is arithmetic and a gap that means something is broken.

This page is about paid campaign measurement: ad platforms, analytics and the system that holds your actual sales. It is the overview. Each cause below has its own diagnosis and its own fix.

The four reasons the numbers cannot agree

Almost every discrepancy you will ever chase is one of four things, or several of them stacked.

Cause What differs Effect you see
Population Who is even eligible to be counted One tool counts everyone, another counts only people who touched an ad
Clock Which day a conversion is filed under Yesterday's numbers keep changing after you screenshot them
Crediting Which touchpoint gets the win Two platforms both claim the same sale
Collection Whether the event was recorded at all A number that is simply too low, everywhere downstream

The first three are permanent. They are design decisions, not bugs, and no tool or tag will close them. The fourth is the only one that is actually a fault, and it is the one worth your afternoon.

Population: the tools are not counting the same people

An ad platform counts an action when the person who took it had previously interacted with an ad from that platform. Analytics counts the action whenever it happens, whoever the visitor is and however they arrived. If any part of your traffic comes from anywhere other than paid ads, the analytics number covers a larger population by definition.

That is before the vocabulary problem. The same event carries different names in different products, and the names have moved: what Google Analytics calls a key event and what Google Ads calls a conversion are two objects with two rulebooks, which we walk through in key events, conversions and the words GA4 keeps renaming. Two people in a meeting using the word "conversion" for two different objects will not agree about anything, and the argument will look like a data problem when it is a definition problem.

The other half of the population question is what you decided to count in the first place. A form submission recorded on a thank-you page and a form submission recorded on a button click are different populations of the same people, as the thank-you page versus button click comparison sets out. One of them counts intentions and one counts outcomes.

Clock: the same conversion, filed under two different days

Google Ads reports its primary conversion columns against the date of the click, not the date the conversion happened. Its own documentation is explicit: "The primary conversion columns mentioned above are calculated based on the time of the click, not the time of the conversion", with the worked example that if the ad was clicked last week and the traffic converted this week, "both the click and the conversion are reported back to last week" (Google Ads Help, "Understand your conversion tracking data", checked 2026-09-07).

Analytics files the event on the day the event occurred. So for any business where people take more than a few hours to decide, the two tools are describing different weeks. The same page offers alternative columns labeled "by conv. time", which report on the date the conversion occurred. Those are the columns to use when you are lining the two systems up, and using them removes a chunk of the gap before you investigate anything.

There is a second clock effect that catches people out. Because credit is assigned backwards, recent days keep filling in after you look at them. A number you exported on Monday is not the number the platform will show for that same Monday two weeks later, and neither version is wrong.

Crediting: two platforms, one sale, two claims

If somebody sees a paid social ad, searches your brand, clicks a search ad and then buys, every platform in that chain can legitimately record a conversion under its own rules. Add them up and you have more conversions than sales. This is the mechanism behind the most common version of the complaint, and it is why the total across platforms will always exceed what your invoicing system shows.

Within Google's own stack, the crediting rules are also not fixed in place. Analytics offers three attribution models, and its documentation notes that "Conversions can be reattributed for up to 7 days after the conversion" (Analytics Help, attribution and attribution modeling, checked 2026-09-07). Credit you already reported can move. Which model to use, and whether changing it would change any decision you actually make, is the subject of attribution models in plain language.

Collection: the only cause that is actually a fault

The first three causes produce a stable gap. Collection failure produces a gap that gets worse, or that appears suddenly, and it is the one to rule out first because it is the only one you can fix.

The common failures are mechanical. A redirect strips the click identifier before the landing page loads, so the ad platform records a click and analytics records a visitor who came from nowhere, which is the gclid problem and the redirect that quietly kills attribution. Untagged links dump paid visits into an unattributed bucket, which is most of what direct traffic in GA4 actually is. Or the conversion exists in one product and never arrives in the other at all, which has its own ordered checklist in why a GA4 conversion is not showing in Google Ads.

The signature to look for: a gap that is roughly steady week to week is structural, and a gap that jumped is a break. Direction helps too. If the ad platform shows conversions and analytics shows almost no matching traffic, suspect the link between the click and the page. If analytics shows the actions and the ad platform shows none, suspect the import or the ad interaction requirement, not the tag.

Which number to believe, by decision

Do not ask which tool is right. Ask which tool's counting rules match the decision in front of you.

The decision Use Why that one
Setting bids or budgets in an ad platform The ad platform's own number It is the number the bidding system consumes. Optimizing toward a figure the platform cannot see does nothing.
Judging a channel against another channel Analytics It is the only one of the three that sees all sources on the same terms.
Reporting revenue or profit upward Your own invoicing or CRM record It is the only system where a number equals money that arrived.
Diagnosing a sudden change All three, compared to their own history A break shows as one instrument moving against its own past, not against the others.

The last row is the rule that saves the most time. Compare a number to itself over time, and never treat one system as the referee of another. A difference between two tools is not a finding. A tool disagreeing with its own last eight weeks, while nothing in the account changed, is a finding.

What to say in the meeting

The reason this becomes an argument is that the slide says "Conversions: 214" with no source and no rules attached. Label the number and the argument mostly stops.

Three habits are enough. Put the source and the clock in the column header, in plain words: "Conversions, Google Ads, click date" and "Key events, GA4, event date". Pick the source of truth per decision before you need it, not in the middle of a disagreement about it. And report one headline number per decision rather than three side by side, because three side by side invites the room to reconcile them, which is a job with no ending.

If someone senior insists on one true number, the honest framing is that the back office number is the one that pays salaries, the platform numbers are steering instruments, and the steering instruments are allowed to read differently from the odometer.

The numbers this page will not give you

Plenty of articles on this query print a percentage: an expected variance, a normal discrepancy range, a threshold above which you should worry. We are not printing one, and the reason is worth stating.

Those figures come from vendor blogs, and most of the pages ranking on this query belong to companies selling an attribution product that the article recommends. We could find no figure for an expected discrepancy attributable to a Google documentation page on 2026-09-07. A range that circulates widely is still a guess, and a guess repeated into your reporting becomes a threshold you make decisions against.

The alternative is better anyway. Measure your own gap for a few weeks while nothing is changing, in the same columns each time, using the "by conv. time" columns where you are comparing to analytics. That baseline is real, it belongs to your account, and it tells you something a borrowed percentage cannot: what normal looks like here.

We also have no first-party audit to cite for this page. The mechanisms above are sourced to Google's own documentation, dated to the day we read it. Both Google Ads and Google Analytics rename features and change defaults on their own schedule, so treat any product-specific statement here as accurate on 2026-09-07 and check it against the linked Help pages before acting on it much later.

FAQ

Should my Google Ads and GA4 numbers match? No, and a correct setup does not make them match. They count different populations, file conversions under different dates and credit touchpoints differently. Expect a stable gap, and investigate when the size of that gap changes rather than because the gap exists.

Which number do I put in the board deck? The one that matches the question being asked. Money and pipeline come from your invoicing or CRM record. Channel comparison comes from analytics. Bidding and paid efficiency come from the ad platform. Label each with its source.

Why do my platforms add up to more conversions than I have sales? Because a single buyer can touch several platforms, and each one can legitimately claim the conversion under its own crediting rules. Totals across platforms overlap and should never be summed.

How do I know if the gap is normal or something is broken? Look at stability and direction. A gap of roughly the same shape week after week is structural. A gap that appeared suddenly, or that keeps widening, points at collection: a stripped click identifier, an untagged link, or a conversion that stopped importing.

Why do last week's numbers keep changing? Google Ads reports primary conversion columns on the click date, so a conversion that happens today can be added to a day in the past. Analytics can also reattribute conversions for up to 7 days after the conversion. Both behaviors are documented and expected.

Sources, read 2026-09-07

Google Ads Help: "Understand your conversion tracking data" (support.google.com/google-ads/answer/6270625). Analytics Help: attribution and attribution modeling (support.google.com/analytics/answer/10596866).

Google Ads and Google Analytics change names, defaults and menus frequently. This page describes mechanisms rather than menu paths for that reason, and carries a six month review date.

Similar Posts