Two weeks ago I taught a masterclass inside Foxwell Founders, a community of about 600 of the sharpest media buyers and ecom operators in the world. These are people who run seven and eight-figure ad accounts for a living.

I didn't teach them a single thing about creative, targeting, or campaign structure. They have that covered better than almost anyone alive.

I taught them a decision system, built around one question every operator eventually faces with real money on the line: how hard can you actually push this account?

And I opened with two real accounts, the same two I'll open with here.

Both are brands we work with. Same question from both founders in the same quarter.

The correct answer for one was the exact opposite of the correct answer for the other, and the ad platforms couldn't tell them apart.

Inside Ads Manager, both looked normal that quarter. Defensible returns, standard creative fatigue, the usual noise.

The numbers that separate them live outside the platforms entirely.

 

Account one: push into a falling ROAS

Early this year, AlgoRX watched their blended ROAS slide from 9.3 toward 4.8. Nearly cut in half.

Every instinct says slow down, and most rooms full of smart operators would have said the same.

The five-number card said push.

Contribution margin after ads sat at 41%, deep green. Cost per new customer was $45 against roughly $90 of first-order gross profit. Payback was covered on the first order. The 90-day ratio cleared 3.

Falling ROAS while every real number stays green has a specific meaning: you're buying growth at a fair price, and the dashboard is just recording that growth got bigger.

So we scaled spend from $83K to $355K a month in five months, in weekly steps of up to 10%, never faster than the numbers allowed.

ROAS kept falling the entire time.

Net profit went from $353K to $667K a month. Up 89%, while the metric everyone else steers by got visibly worse.

The rule underneath: scaling is only real if CM3 holds while spend rises. Theirs held at 41 to 44% through a 4.3x spend increase.

If spend is up and CM3 is down, you're buying revenue, not building a business.

The tool behind that climb is worth stealing.

It's one slide, an allowable-CAC ladder computed straight from their margins: breakeven at $90 per new customer, the 10%-net-margin line at $73, the 20%-net line at $57.

As spend climbed, their real nCAC walked from $45 to about $80, and it never crossed the breakeven rung. Every weekly budget decision was a glance at a ladder instead of a debate.

 

Account two: stop cold

Gnarly Nutrition arrived with fresh investor money in the bank and a plan to pour it into marketing. Growth mode, foot down.

Same card, opposite verdict.

Contribution margin after ads was deeply negative. Cost per new customer ran $138 to $144 against roughly $79 of 90-day customer value. Payback sat around eight months.

Scaling that account would have meant scaling losses with investor money, faster every month.

And to be clear, nobody inside that brand was doing anything dumb. The creative was fine, the media buying was competent, and the plan looked like every growth plan you've ever seen.

The economics underneath simply couldn't carry paid acquisition yet, and no amount of optimization inside the ad account could outrun that.

So the call was stop. Ad spend got cut to about 30% of the prior year.

A few months later the brand hit its first profitable stretch in over a decade, and new-customer revenue is now growing on a fraction of the old budget.

The line I gave the room: the card's job isn't to let you scale. Its job is to tell you the truth.

 

The system itself

Five numbers, each with hard thresholds:

• Contribution margin after ads (CM3)
• New-customer CAC against first-order profit
• The 90-day LTGP-to-CAC ratio
• Payback period
• New-versus-returning revenue split

Then one decision rule on top:

• All five green: scale 20 to 30% a week
• One yellow: 10 to 15%
• Two yellows: hold and fix
• CM3 or payback red: stop immediately

And the discipline that makes it work: your worst metric on the card is your constraint, and it's the only thing worth touching this week.

Fixing a green metric does nothing. Fixing the red one raises the ceiling on the whole account.

Notice what's missing. ROAS is not on the card.

That's not an oversight. Both stories above are the reason why.

All five numbers come from your store backend and your P&L. None of them exist inside Meta or Google, which is exactly why two accounts that look identical in the platform can need opposite decisions.

The platform sees clicks and conversions. The card sees whether they're worth anything.

 

Where operators go blank

I had the room score their own accounts from memory, one letter per metric: G, Y, R, or a question mark.

Here's what 417 recorded founder calls have taught me about that exercise: almost nobody blanks on CM3 or CAC. The question marks cluster on payback and the 90-day ratio, every single time, at every level of sophistication.

Those two numbers are where "I know my numbers" turns out to mean "I know some of my numbers."

The thing I repeated three times that hour: sophisticated operators don't optimize more things. They know their real numbers, and they fix one constraint at a time.

 

Your turn

Score your own account right now, from memory, five letters.

Every question mark is homework, and most five-letter strings I see have at least one. Your worst color is your constraint, and the only project that matters this week.

And if you drew question marks, no shame in it. Payback and the 90-day ratio need cohort views your dashboards don't surface by default, so nobody trips over them by accident.

You have to go get them on purpose, and almost nobody has. That's exactly why they're worth having.

The part that stuck with me came in the days after the call.

Operators started reaching out, and the messages that meant the most had nothing to do with the presentation. They'd gone and run the card on their own accounts, and more than one landed on a version of the same sentence: I finally know what my actual constraint is.

That sentence is exactly what the card is for.

The same scorecard I gave the room, with the formulas, thresholds, and verdict logic built in, is free here.

Real numbers. One constraint. That's the whole system.

It held up in front of that room. It'll hold up in your account tonight.

Talk soon,
Andrej

PS: I made the room an offer and I'll make you the same one. Reply "teardown" with your five numbers, or your five letters, and I'll run the card on your account and tell you your one constraint. Free, no pitch. I'll do as many as I can.

PPS: The two ways to set next week's budget →