Hygiene.

Nobody had asked for an audit.

The call was a repair job.

The company sold peptides, and it spent heavily on Meta ads.

Its agency had a problem every busy ad account eventually has.

The ad platform knew what every ad did. How much it spent. How much it brought back.

A separate project board knew what every ad was. Who it was meant for. What buying motivation it was built around. What format it used.

Two systems. Two halves of the same story. And nothing connecting them.

Andrew had built a small app to fix that. It pulled each ad’s performance into the matching creative record on the board, so the team could finally see results and intentions side by side.

The app had been throwing errors. So late one night, Andrew got on a call with the agency operator responsible for the client, to get it working again and set him up so he could manage it himself.

For most of an hour, that’s exactly what they did.

Connections. Access tokens. Remote access. A second brand that needed its own setup.

The unglamorous plumbing that makes everything else possible.

Once the connection was working, Andrew wanted to make sure the data coming through was real.

So he pointed an AI analysis at the account.

Not a formal engagement. A test.

Almost immediately, it flagged two ads that were "burning ad spend that’s going nowhere."

The operator asked what that meant.

The operator had a worry about this kind of thing. He’d been reading about people getting their ad accounts shut down for wiring AI directly into them.

Andrew wasn’t talking about letting AI run the account.

Then the first numbers came back.

A 77% checkout abandonment rate. A 30% drop between someone clicking an ad and the page actually loading. A $178 cost to acquire a customer worth $462 on average.

Andrew looked at that last pair and asked the obvious question.

Remember that.

Two people on the call. Neither of them the person whose job it was to optimize this account.

And they were about to find out what that person had been missing.

About an hour and seventeen minutes in, the full analysis came back.

Andrew started reading.

There was a group of ads being fed around $8,000 a week, every one of them returning less than $1.50 for each dollar spent.

Ads running between one and seven in the morning were losing about $13,000 a week, returning less than a dollar for every dollar.

Meanwhile, a set of ads returning between 3.5 and 6.8 times their spend were being starved. The account wasn’t giving them enough budget to do what they were already proving they could do.

And then the big one.

Of 215 ads, 158 had returned less than 1.5 times their spend over the previous seven days.

Seventy-five percent of the account.

Together they had spent about $45,000.

They had brought back about $36,000.

Maybe you’ve seen some version of that in your own business. Not one disastrous decision, just a long tail of small ones that nobody went back to check. Each ad looked fine on the day it launched. Nobody added up what all of them were doing together.

The analysis was blunt about what came next. Pause the tail. Move that money to the ads that were already working. At the same total spend, it estimated, that alone would add around $30,000 a week in revenue.

Andrew read the line that summed it up.

The operator saw something bigger.

The findings kept coming.

The account was, in the analysis’s words, "one ad deep." A single ad was driving 28% of the revenue.

The landing page was quietly losing almost a third of every paid click before anyone saw the cart.

The ads were paying for video, but people were clicking before the video even played. The average watch was two seconds.

And two of the account’s top performers had been paused in the middle of the week. Andrew didn’t soften it.

At this point, the operator did what a careful operator should do.

Fair question. He wanted to look at it fresh in the morning.

Then he described what he really wanted. An AI analyst that didn’t just read the ad platform, but read the creative records too. Buying motivation. Persona. Format.

Because that had been the whole point of the project board.

The client’s creative team used to make random ads, he explained, and then hit ceilings because everything started to look the same. The board was supposed to fix that. Every ad would carry its reasons with it, so when something worked, the team would know why, and could make more of it.

That was the system.

Andrew had the report open in front of him.

An hour and thirty-five minutes in, Andrew reached the last finding.

More than a third of the money going out the door couldn’t be connected back to any creative record.

And it got more specific.

Then the second.

Sit with that for a second.

The entire reason the project board existed was to capture why winning ads won, so the team could repeat it.

And the two best ads in the account were exactly the two it couldn’t explain.

The operator’s skepticism didn’t last long after that.

Andrew didn’t walk into that call as a media buyer.

He said so himself.

He walked in to fix a pipe between two systems.

But once the pipe was working, he did something most people skip. He checked what was actually flowing through it.

And the most valuable thing he found wasn’t the $45,000.

Plenty of people can eventually spot a tail of underperforming ads.

It was the gap between the two systems. The place where one team could see the results and another team could see the reasons, and nobody could see both. That gap had swallowed the account’s two best ads.

That’s the kind of problem that hides between departments, between tools, and between people who each assume the other one is watching.

The fix for the first problem was hygiene.

The fix for the second was making sure the next winning ad came with its reasons attached.

Andrew had built the pipe for that.

He just hadn’t expected it to show everyone how badly they needed it.