Which Marketing Numbers Should You Trust When Google Ads, Meta, GA4 and Your CRM Disagree?

Why marketing systems disagree

Some disagreement is normal.

Google documents multiple reasons Google Ads and Analytics can show different conversion totals, including reporting timing, conversion windows, counting settings and attribution settings. Google Analytics can also use different eligible attribution channels and reporting time zones from Google Ads. Google Help Google Help

Source:Google Help

Source:Google Help

Add Meta, CRM records, ecommerce systems, offline sales and identity limitations, and exact reconciliation becomes even less realistic.

But “the numbers never match” is not a diagnosis.

Some discrepancies are expected. Some mean you are comparing different outcomes. Others indicate a measurement failure serious enough to invalidate a decision.

The first job is determining which situation you have.

Related diagnostic:Why Does My ROAS Look Good but Revenue Isn't Growing?

The TAC Measurement Disagreement Protocol

1. Decision: What are you actually trying to decide?

Start with the decision.

Not:

The Agency Check

Which dashboard is correct?

Instead:

The Agency Check

What decision are we trying to make?

You might be deciding:

• whether one Google Ads campaign should receive more budget;
• whether one Meta creative is outperforming another;
• how many qualified opportunities marketing produced;
• which channels appear to contribute to pipeline;
• whether an executive marketing report is representing business performance accurately.

Those are different decisions.

They do not automatically require the same evidence.

That is why you choose the decision before you choose the number.

2. Outcome: Are the systems measuring the same thing?

Before debating attribution, define the outcome.

Imagine the dashboards show:

Those numbers may not contradict one another.

Google Ads may be counting submitted forms.

Meta may be reporting attributed lead events.

GA4 may be recording website form activity.

The CRM may be showing people who progressed far enough to become genuine sales opportunities.

If so, 184, 96, 207 and 41 describe different stages of the funnel.

Trying to reconcile them as though they measure the same outcome creates a problem that does not exist.

Before comparing two numbers, ask:

The Agency Check

What event is each system actually counting?

3. Business record: What do we reliably know actually happened?

Next, establish the closest reliable record of the business outcome.

For ecommerce, that might be completed orders, net revenue or customers recorded by commerce and finance systems.

For lead generation, it might be:

Lead → Qualified lead → Opportunity → Customer → Revenue: The CRM can therefore be extremely important.

But “trust the CRM” is still too simplistic.

If sales representatives do not update opportunity stages, duplicate records are common, offline deals are missing or acquisition data disappears during handoff, the CRM may contain valuable business outcomes while remaining unreliable for a particular marketing question.

Do not ask only:

The Agency Check

What is our source of truth?

Ask:

The Agency Check

What do we reliably know actually happened, and where is that outcome recorded?

That forces the underlying data quality to earn your trust.

4. Disagreement type: Why do the numbers differ?

TAC separates measurement disagreement into four types:

These require different responses.

You do not repair legitimate attribution disagreement as though it were a broken tracking tag.

And you should not dismiss a genuine tracking failure by saying attribution is imperfect.

First identify what kind of disagreement you are looking at.

5. Decision rights: Which source should govern this decision?

Now assign decision rights.

In plain English, decision rights means:

The Agency Check

Which source gets to govern this specific decision?

Not which system is universally “correct.”

Which source has the most appropriate and reliable evidence for the decision being made?

The word reliable matters.

A broken Google Ads conversion action should not govern bidding.

A badly maintained CRM should not govern channel comparisons simply because it is the CRM.

The source earns decision rights by being fit for that specific decision.

6. Decision status: Is the evidence good enough to act?

Finish by classifying the decision.

Valid: The relevant evidence is sufficiently reliable.

Valid with limitations: The decision can still be made, but known uncertainty needs to be acknowledged.

Decision-breaking: The measurement problem undermines the evidence required for the decision. Stop making that decision until measurement is repaired or a defensible alternative is established.

Not every discrepancy deserves an analytics emergency.

But some absolutely do.

Related Google Ads diagnostic:How Can I Tell If My Google Ads Are Actually Being Optimized?

What does a decision-breaking discrepancy look like?

Imagine a B2B company wants to compare cost per qualified opportunity by paid channel so it can reallocate next month's advertising budget.

Then the team discovers that a recent CRM migration lost original-source data for a large share of new opportunities.

Google offline conversions are still importing for some records.

Meta-qualified outcomes have stopped syncing.

The business still has real spend and real opportunities.

But its ability to connect those opportunities consistently back to paid channels is broken.

The company can still know:

• how many opportunities exist overall;
• how much each channel spent;
• what activity occurred inside each advertising platform.

What it cannot reliably calculate is comparable channel-level cost per qualified opportunity.

The correct action is not:

The Agency Check

Google looks better, so move more money there.

It is:

The Agency Check

Stop using channel-level qualified-opportunity economics to reallocate budget until the CRM/source mapping is repaired or another defensible measurement method is established.

That is decision-breaking.

The problem is not that every number is useless.

The problem is that the evidence required for that particular decision is unreliable.

A fully worked example: Google Ads, Meta, GA4 and CRM disagree

Consider an illustrative B2B company running Google Ads and Meta Ads.

These numbers are examples, not TAC client data.

At month end, the team sees:

Management asks:

The Agency Check

Which paid channel is producing more qualified opportunities?

Run the protocol.

Step 1 — Decision

The actual decision is not:

The Agency Check

Which platform reports more conversions?

It is:

The Agency Check

Which paid channel appears to be producing more sales-qualified opportunities?

That requires downstream quality information.

Platform conversion totals immediately become supporting evidence rather than the final answer.

Step 2 — Outcome

The systems are not measuring the same outcome.

Google Ads is optimizing toward a submitted lead form.

Meta reports attributed lead events.

GA4 records the website form event.

The CRM records people and then separately identifies which ones sales accepts as qualified opportunities.

The company has been comparing:

marketing events with sales outcomes.: That is a definition difference.

Step 3 — Business record

The CRM contains 32 opportunities that sales has accepted as qualified, and a sample review confirms those stages are being maintained consistently.

For the question “how many qualified opportunities exist?”, the CRM therefore has strong evidence.

But source attribution is less certain.

Some prospects interacted with more than one paid channel. Some clicked Google after first encountering Meta. Others have incomplete acquisition-source data.

The qualification data is reliable.

The precise channel assignment is less reliable.

Step 4 — Disagreement type

Several disagreement types exist at once.

Google/Meta versus CRM is mainly a definition difference because platform lead conversions are not sales-qualified opportunities.

Google versus Meta includes an attribution difference because both platforms may claim influence over overlapping journeys.

GA4 versus CRM includes an observation difference because GA4 sees website behavior while the CRM sees what happened later in the sales process.

A small discrepancy between GA4 form events and CRM lead records is investigated and traced to test submissions and duplicates rather than a major tracking failure.

There is no single discrepancy to “fix.”

There are several explainable differences.

Step 5 — Decision rights

For the number of actual qualified opportunities, the reliable CRM record gets primary decision rights.

Google Ads still needs the right conversion signals to optimize Google campaigns.

Meta's own relevant signals remain useful for Meta campaign and creative decisions.

GA4 remains valuable for website behavior, landing pages and acquisition journeys.

But none of those platforms automatically gets the right to declare:

The Agency Check

We produced X qualified opportunities.

unless the qualification outcome can actually be connected back reliably.

Step 6 — Decision status

The company can confidently conclude:

The Agency Check

We generated 32 sales-qualified opportunities.

It can also evaluate raw lead generation and campaign behavior inside each advertising platform.

However, because some qualified opportunities have incomplete or overlapping acquisition histories, management cannot confidently claim that one paid channel caused a precise share of those 32 opportunities.

The decision is therefore:

The Agency Check

Valid with limitations.

There is enough evidence to make some optimization decisions.

There is not enough evidence to support every attribution claim management originally wanted to make.

The TAC Measurement Decision Record

The example can now be reduced to one decision record:

One important decision gets one record.

This prevents the business from silently changing which number it trusts depending on which dashboard supports the preferred conclusion.

Watch how the disagreement changes, not only how large it is

A known discrepancy can become a useful monitoring signal.

Suppose Meta has consistently reported more conversions than GA4.

That may be explainable.

Then the normal relationship changes sharply.

That change deserves investigation.

The signal is no longer simply:

The Agency Check

Meta and GA4 disagree.

It is:

The Agency Check

Their normal relationship changed.

That may point to altered conversion settings, duplicated or missing events, tracking changes, consent effects, CRM problems or a genuine change in customer behavior.

Google similarly recommends reviewing tagging, conversion settings, reporting timing and attribution when discrepancies appear. Google Help

Source:Google Help

A useful measurement system therefore monitors both the individual numbers and the relationship between them.

Do not force legitimate differences to reconcile

Making every dashboard show the same conversion total is not the goal.

Sometimes the systems genuinely answer different questions.

Google Ads may be supporting Google's campaign optimization.

GA4 may be helping explain website behavior and cross-channel journeys.

The CRM may be verifying downstream sales outcomes.

Perfect numerical agreement is neither necessary nor always possible.

The stronger standard is:

The Agency Check

Make every important decision traceable to evidence appropriate for that decision.

What should your agency or internal team be able to explain?

If a report contains materially different numbers across systems, whoever owns measurement should be able to explain:

• what each important number represents;
• why the systems differ enough to matter;
• which source is governing the decision, and why.

TAC's existing guide to What Should a Marketing Agency Report Include Every Month? already argues that good reporting should distinguish what is known, estimated and uncertain, and identify which source is being used for decision-making. The live article also explicitly addresses disagreement between advertising platforms, analytics and CRM data. The Agency Check

Related reading:What Should a Marketing Agency Report Include Every Month?

This framework goes one level deeper:

The Agency Check

Does that source actually deserve to govern the decision?

For internal teams:Independent support for in-house marketing teams

For companies working with an agency:Independent review of your agency's marketing performance

This does not answer whether your total marketing budget is working

This boundary matters.

The TAC Measurement Disagreement Protocol answers:

The Agency Check

Can the available evidence validly support this particular marketing decision?

It does not answer:

The Agency Check

Is the company's total marketing investment creating enough business value?

Those are separate analyses.

You can have excellent measurement and discover that overall marketing economics are weak.

You can also have strong business growth while attribution between individual channels remains uncertain.

Measurement validity comes first because unreliable evidence can undermine the business-value analysis.

Once the evidence is sufficiently trustworthy, the next question becomes economic:

The Agency Check

Are we receiving enough business value from what we are spending?

That is a different diagnostic.

When should you investigate immediately?

A discrepancy deserves immediate attention when it threatens the decision.

Examples include:

• systems that should be counting the same event suddenly diverging;
• reported outcomes exceeding what the business could actually have produced;
• missing or duplicated conversions;
• broken CRM imports;
• an unexpected change in tracking configuration;
• evidence required for an important decision becoming incomplete.

Google provides dedicated conversion diagnostics because broken or misconfigured tracking can affect both reporting and campaign optimization. Google Help

Source:Google Help

The standard is simple:

The Agency Check

Expected difference: explain it. Material uncertainty: disclose it. Measurement failure: repair it. Decision-breaking failure: stop the affected decision until the evidence is reliable enough to continue.

The goal is not one source of truth. It is a trustworthy decision system.

Marketing measurement becomes easier to manage when you stop asking every platform to tell the same story.

Google Ads needs reliable evidence for Google Ads decisions.

Meta needs useful evidence for Meta decisions.

Analytics needs to help explain website behavior and acquisition journeys.

CRM, commerce and finance systems need to record meaningful downstream outcomes reliably.

And management needs to know when none of those systems can support the conclusion being requested.

The Agency Check

Choose the decision before you choose the number.

Then make sure the number has earned the right to influence that decision.