What If Your Biggest Orders Are Teaching Meta to Find the Wrong Customers?

“Meta's like, well, it was this ad that did it.”

Her reaction?

“Oh, no. Like, I don't want you to go after that type of person.”

Because she could see exactly what was happening.

Whenever a bulk purchase came through, her average order value would shoot up.

She knew that wasn't a normal customer.

So she could spot it.

She could pull it out.

She could manually recalculate her ROAS without it.

But that only fixed the number she was looking at.

It didn't fix what Meta was learning.

As she put it:

“I can manually pull it out to calculate the true ROAS… but it's still feeding that into the algorithm.”

And that's where an annoying attribution problem can turn into something much more expensive.

Because Meta isn't just reporting on your conversions.

It's learning from them.

Every purchase becomes another signal that helps the algorithm decide which ads are working, what a valuable customer looks like and where it should put more of your money.

So when a huge bulk purchase gets credited to an ad…

Meta doesn't know you're sitting on the other side of the screen saying:

“No, don't.”

It sees a win.

A BIG one.

And does exactly what an optimization algorithm is supposed to do.

It tries to find more.

More people like that buyer.

More conversions like that purchase.

More opportunities to recreate the outcome that just made an ad look incredibly valuable.

There's just one problem.

That's not the outcome you wanted it learning from.

For this marketing leader, the customers she really wanted to understand were subscription buyers.

She was actively testing different creative themes to figure out what made those people subscribe.

Was it convenience?

Softness?

Being clean and non-toxic?

She wanted to know which message was actually bringing in the right customer so she could build more creative around it.

But those bulk purchases were getting mixed into the signal.

And suddenly an ad could look like a monster winner for the wrong reason.

That's the trap.

You open Meta.

ROAS looks fantastic.

Revenue looks fantastic.

One ad appears to be crushing everything else.

The obvious conclusion?

Give the winner more money.

Meanwhile, the “winner” may be teaching Meta to get better and better at finding a completely different type of buyer than the one you're actually trying to acquire.

And she had already tried to stop it.

She built a custom conversion.

She pulled in the product IDs for the bulk products so they could be excluded.

Then she thought she'd finally solved it.

“I thought it was working.”

Until another heavy bulk-order week came through.

And then?

“Every single one of them” got credit.

Her response was simple:

“No, don't.”

That's what makes this problem so easy to underestimate.

Most experienced marketers already know the number inside Meta isn't gospel.

You can sanity-check it.

You can identify an outlier.

You can pull a bulk purchase out of your calculations.

You can explain the weird ROAS spike to leadership.

But there's a much harder problem hiding underneath all of that.

You can correct the number you see without correcting the signal Meta sees.

And if the signal is wrong, the algorithm can become extremely good at optimizing toward something you never wanted in the first place.

That's why the better question isn't always:

“Which ad has the highest ROAS?”

Sometimes it's:

“What is Meta learning from the customers it's giving this ad credit for?”

Because this same marketing leader had already started seeing what happens when you stop relying on each platform's version of the truth.

According to Meta, her ROAS was declining.

According to GA4, it was rising.

And after making optimizations using an independent view of performance, Google, historically an underperformer for her, started performing really well.

Subscriptions climbed higher than she'd seen since joining the company.

So she started changing the question entirely.

Instead of:

What did Meta do?

What did Google do?

What did each platform say its ROAS was?

She started asking:

“How about just overall, how effectively did we spend our money?”

That's the shift.

Because your ad platforms are very good at optimizing toward the signals they're given.

But they don't know your business like you do.

They don't know which unusual order you want excluded.

They don't know which customer you're actually trying to acquire.

And they don't know when a giant “win” is teaching them exactly the wrong lesson.

That's why sophisticated marketing teams are starting to build an independent view of performance before deciding what deserves the next dollar.

One that lets them see how channels are working together.

Which ads are actually influencing the customers they care about.

And whether the “winner” Meta wants to scale is actually helping create the growth they want.

Because sometimes the most dangerous ad in your account isn't the obvious loser.

It's the “winner” that's teaching Meta to win the wrong game.