When Your New Agency Says Your Marketing Numbers Are All Wrong... And Turns Out They’re Right

“There’s no way this is what we’re doing on Meta.”

The new agency wasn’t tiptoeing around it.

“Your reporting is off.”

“These numbers are off.”

They had just started working with the company. And on one of their first calls with us, they were already aggressively questioning the marketing numbers in front of everyone.

There was a good reason.

For one ad set, Meta was reporting roughly 3X the conversions showing up in our GA4-based reporting.

Meta said 38.

GA4 showed 12.

That kind of gap changes the question.

It’s no longer:

Which attribution model do we prefer?

It’s:

What the hell is actually happening here?

So we investigated.

And when we opened their Meta environment, we found something nobody had been looking for.

Seven active pixels.

Some were backups. One came from an old agency.

And the same purchase event was landing in at least three of them.

One purchase could effectively show up multiple times.

Which meant the problem wasn't that Meta was making their marketing look worse.

It was making their marketing look better.

And that can be much more expensive.

Because when marketing looks bad, someone asks questions.

When marketing looks great?

You scale it.

An inflated conversion count makes ROAS look stronger.

It makes CPA look lower.

It makes an ad set look like a winner.

And eventually somebody looks at those numbers and decides:

Put more money there.

That's how a tracking problem can become an allocation problem.

The dashboard doesn't have to fail spectacularly.

It only has to make the wrong opportunity look convincing enough to act on.

Why problems like this can hide in plain sight

Nobody deliberately builds a measurement environment designed to triple-count purchases.

It accumulates.

An old agency installs a pixel.

Someone creates another as a backup.

A new implementation gets added.

Another gets created to verify the previous setup.

Each decision can make sense in isolation.

Eventually you have several systems collecting versions of the same event, and nobody has a clean picture of everything feeding the number the marketing team is using to make decisions.

That's essentially what happened here.

And ironically, during the investigation, the instinct was to create another custom event to cross-check the existing numbers.

It sounds reasonable.

When you don't trust one number, get another number.

But adding another measurement layer before understanding the existing ones can make the environment even harder to untangle.

So instead of adding another number, we traced the one that was already there.

Same ad set.

Same date range.

Same purchase event.

Then we followed where that event was actually going.

That's what exposed the multiple active pixels.

The number isn't the problem. The mechanism behind the number is.

This is an important distinction.

When two marketing systems disagree, the natural reaction is to ask:

Which number should I trust?

Sometimes that's the wrong first question.

Because choosing one dashboard over another doesn't explain the discrepancy.

You have to understand why the numbers disagree.

In this case, the direction of the discrepancy itself was a clue.

Our attribution window was longer than Meta's.

All else equal, you would expect the longer window to uncover more conversions.

Instead, Meta was showing roughly three times as many.

That made a simple attribution-window explanation difficult to reconcile with what we were seeing.

So we kept tracing the mechanism.

And eventually the numbers stopped being mysterious.

The same purchase event was reaching multiple active pixels.

Once you see that, the discrepancy isn't an attribution debate anymore.

It's a data problem you can actually fix.

That's the kind of problem you want your measurement system to expose

There’s a larger lesson here for marketing leaders.

The goal isn't to make every dashboard agree.

And it certainly isn't to choose whichever number makes performance look best.

The goal is to understand your marketing well enough that when the numbers don't make sense, you can trace the disagreement back to what is actually happening.

That means connecting the marketing activity you're seeing in platforms like Meta to an independent view of the downstream business events you actually care about.

Purchases.

Leads.

MQLs.

SQLs.

Subscriptions.

Revenue.

Blueprint starts by getting that underlying data environment right, then creates different views of how marketing contributes to those outcomes.

What can be directly attributed.

What influenced the customer journey.

And what appears to be having a broader impact on the business even when a direct click can't explain it.

That layered view matters because no single platform sees the entire customer journey.

And the platform reporting the performance also has to make decisions with the information available inside its own environment.

The marketing executive has a different job.

You have to decide where the next dollar goes.

That requires more than accepting the number sitting inside the channel reporting.

It requires understanding what produced it.

In this case, the agency did exactly what you'd want a good agency to do

They saw something that didn't make sense.

And they aggressively called it out.

We investigated it.

The client's team made the changes.

And the result wasn't another dashboard everyone had to learn to tolerate.

It was a much clearer understanding of what their Meta performance actually looked like.

That changes the decision the marketing leader can make.

Instead of:

Meta says this ad is crushing it. Should we scale it?

You can start asking:

Is this ad actually creating the business outcome we think it is?

What else contributed to those customers?

Does the apparent opportunity survive when we look beyond Meta's own reporting?

And ultimately:

Is this really where our next dollar should go?

Because the dangerous marketing number isn't always the one that looks obviously broken.

Sometimes it's the number that looks so good...

nobody thinks to question it.

FAQ

Why can Meta report more conversions than GA4?

Meta and GA4 use different attribution methodologies, so some disagreement is normal. But a large discrepancy can also indicate a structural tracking issue. In this case, multiple active pixels were receiving the same purchase event, which inflated the conversion reporting inside Meta.

Can multiple Meta pixels cause duplicate conversions?

Yes. Separate pixels or datasets can receive the same purchase event as separate data streams. In this client's setup, the same purchase event was landing in multiple active pixels, contributing to inflated reporting.

Why are inflated conversions more dangerous than missing conversions?

Missing conversions can cause marketers to undervalue something that's working. Inflated conversions create the opposite risk: an ad or channel can appear more efficient than it actually is, encouraging the team to put additional budget behind an opportunity that isn't real.

What should you do when Meta and GA4 disagree significantly?

Don't immediately choose one number as the source of truth. First investigate the mechanism behind the discrepancy: compare equivalent time periods and campaigns, trace the conversion event, review the active pixel and dataset environment, and determine where the same event may be entering the measurement stack more than once. That's the process that exposed the problem in this case.

Should you use platform-reported ROAS to decide where to scale?

Platform-reported ROAS can provide useful context, but it should not automatically determine allocation decisions. A marketing leader needs to understand how channel activity connects to downstream business outcomes and whether the underlying tracking supports the apparent performance before putting substantially more money behind it.