If you overspend on losers, it drags the entire account. Most people don't have a system that stops that from happening.

Today I'm breaking down the structure we've been running, why it's built the way it is, and what happens as you move a creative from test to scale using this system.

Most people running at scale have two campaigns: a testing campaign and a scale campaign.

Direct promotion from test to scale is a coin flip. A creative that looks great on a $25/day test gets dropped into a campaign spending $15k or more a day, and one of two things happens. It either eats the whole budget before you know if it converts at volume, dragging the entire account down with it.

Or it never builds enough estimated action rate to take meaningful spend, so it sits there doing nothing while you wonder why your winner died. Either way you, lose. The fix isn’t a better test, but adding a new layer between.

This is the structure we run on accounts above $25,000 a day. Four campaigns, each with one job.

Starting budget: $25 a day per ad set

One ad set, two ads per ad set, 80+ creatives loaded in. The whole thing runs on automated rules.

Above your CPA target, the rule scales an ad's budget down 10%. Below it, the rule scales it up 10%. The ads manage their own budgets. No manual babysitting, no accidentally killing a winner at 9am because the morning numbers looked soft.

The point is maximum creative volume without overspending on losers, because the losers throttle themselves down automatically.

In creative testing we leave these creatives on and continue to let them do their thing.

Once a creative consistently performs well, it earns a "V" for Validated and we move it to a small ASC campaign running around $3,000–$4,000 a day.

This is the part that makes everything else work, and it's the part that most accounts skip. The job of this campaign is signal, nothing else.

A creative that won on $25/day hasn't actually proven anything. It hasn't been given enough spend for Meta to build a reliable estimated action rate, or EAR, which is the auction's read on how likely someone is to take the action you want.

Validation feeds the creative enough data to build that EAR inside a controlled budget. You wait five to six days. The real winners float to the top and keep performing below your CPA target. When one does, it gets marked with an S and sent to scale. We kill the creatives which are moved from the validation campaign.

The ones that don't float never get near your main budget. That's the whole point.

Everything that survived validation lives here, and this is where creative actually scales. Budget is $15,000 to $20,000 a day. It's a CBO, and the ad sets are organized by month: June winners, July winners, August winners. Winners from each month stay together.

By the time a creative reaches this campaign, it has already proven it converts and already carries the EAR to take spend without starving everything around it. No coin flip. You only ever scale things that earned their way up the ladder.

Partnerships run on their own track. Same automated rules as creative testing, just isolated in a separate campaign on an ABO setup.

Each one starts at $25 a day, raises 10% when it's below CPA, scales down when it's above. So at any moment some are climbing and some are walking themselves back. A partner sitting at $99.72 against a $100 target keeps scaling. One above $100 throttles down on its own.

Same scale-in-place logic, kept separate so partnership spend never muddies your creative testing data.

Two of these campaigns scale in place on autopilot (testing and partnerships), feeding winners and throttling losers without anyone touching them. The other two (validation and main scale) move proven creatives up one earned step at a time.

The entire structure exists to kill one specific failure: the creative that looks like a winner on a tiny budget and then either bankrupts your scale campaign or flatlines inside it. You stop guessing which is which, and let the system tell you.

I’ve been looking forward to sharing this strategy for a while now - hopefully you can use it to get some amazing numbers. It’s been absolutely crushing for our clients.

The 10% rule is deliberately small. Big budget swings spook the algorithm and reset learning. Moving in 10% steps lets winners climb and losers fade without tripping the campaign back into the learning phase.

Two ads per ad set in testing isn't random. It gives each creative a fair shot at delivery without splitting a tiny test budget across so many ads that none of them get enough impressions to mean anything.

Organizing scale ad sets by month does more than keep things tidy. It shows you creative decay in real time. When June's ad set starts sliding while July's holds, you can watch your winners age before the account-level numbers ever tell you.

This is the exact structure we run on six-figure-a-month Meta accounts every day. If you're spending at scale and the test-to-scale jump still feels like a gamble, book a call and we'll look at how your account is built.

— Torii, founder @ DREAMLABS

P.S. The most common mistake I see at scale is promoting a creative to the big-budget campaign the day it wins in testing. If you take one thing from this issue, make it the layer in between.