READ THIS FIRST

The old testing best practices are invalidated

Meta Andromeda exists now, and it changed what testing is for, when it makes sense, and when it is actively costing you money.

This is the structure running right now inside agencies that spend heavily month in and month out, very profitably, on deals they hold net profit shares in.

Who this is for

Ideally you are already doing a couple hundred grand a month, or you have passed a million a month and want to do it again.

The perspective

Everything here is written from a scaled point of view, not a first-campaign one.

THE TIMELINE

Why everything changed

This is not a preview of something coming. It is already on your account.

If you are reading this, with an extremely high degree of certainty it has already rolled out on your account. If you are not updated to this set of best practices, you are getting absolutely cooked.

This playbook is not about Andromeda specifics. It stays focused on creative testing.

CAMPAIGNS

Two campaigns, three ad sets

We typically run two campaign types: cold and warm. We barely run traffic to warm audiences unless the audience is big enough to be worth real dollars. You can always run it; it will just be a very low-budget campaign.

Broad
An interest stack: a lot of interests, behaviors and demographic traits loaded into one giant ad set, not separated out.
A lookalike stack: a bunch of 1% lookalikes off things like highest-value customers, customers, leads that showed up to calls, and qualified calls that booked. Use whatever else you find valuable.

All three ad sets contain the exact same ads. Broader is usually better now.

TARGETING

Messaging controls targeting now

This trips people up, because it is the opposite of how targeting used to work. Targeting is rarely controlled at the ad set level anymore. Messaging controls targeting now. That is why everything can go into one broad stack instead of slicing audiences into pieces.

Setting	What to do
Budget	ABONot CBO, so you can control distribution across ad sets a little
AI Advantage creative enhancements	OffAll of them, including site links (the “additional links” option)
Highlight a positive comment	OnThe only enhancement worth keeping right now

You rarely choose your audience at the ad set level anymore. You choose it with what you say.

DIVERSIFY

25 to 30 reasons, not 25 to 30 hooks

Andromeda rewards messaging diversification. So create 25 to 30 individual reasons somebody should convert on your offer, and put each one into its own ad.

Not this

One body and CTA with 25 different hooks attached. A new hook on the same ad is not a new reason.

This

25 to 30 unique creatives, each carrying its own specific reason to convert. Video ads, image ads, or a mix of both.

The creative has to actually be different.

FOR THE DETAIL PEOPLE

Five body copies, five headlines, shared

Alongside the 25 to 30 unique ads, write five pieces of body copy and five headlines.

For everyone about to tweak out: those are the same five headlines and five body copies for every single ad, written broad enough to apply across the board. Writing five of each for every individual ad would be exhausting.

TRY IT

One to three ads will eat all the reach

About one to three ads get all the reach and all the distribution. Launch with 30, and 27 to 29 of them end up with little to no reach. They spend pennies, maybe dollars if you are lucky.

Ready to launch. Thirty ads, none of them running yet.

Different ads win each time, and the spend still lands on just a few. That concentration is the system working, not your test failing.

THE OLD BELIEF

“Every creative deserves distribution”

You will feel like it was wasted effort when you launch 25 to 30 ads and one to three get all the reach. You will look at the videos that got nothing and think this is terrible. That belief is what gets you cooked, because of what you do next: put one ad in one ad set, and duplicate that ad set 25 to 30 times so every ad gets its own distribution.

An insane level of audience overlap. You hammer the same people across 30 ad sets.
Artificially inflated costs. You pay more and get less, every time it gets tested.
You fight the algorithm. You force distribution onto creatives a trillion-dollar company has already decided are less likely to convert.
You break account simplification, which Meta still rewards.

You think you are running a creative test. You are running an overlap problem with a bigger bill.

BEING FAIR

Being right doesn't matter. Making money does.

Are there cases where the one-ad-per-ad-set way still works? Yes. And if it started working better, the right move would be to switch. The point is not to be right; it is to make money.

This structure gets tested against the alternative across many accounts and live deals, to prove it in reality rather than in theory. It wins.

And the ads that never got used after you spent time filming them? That is not a problem at all. It is a good thing. You will see exactly why in the next act.

THE LOOP

Test your way out of testing

Out of the 25 to 30 ads, one to three get determined as winners. When you get that combination of winners, you start to scale it. Most people never really do this. They get stuck testing their way to poverty.

Test, scale, find the ceiling, duplicate, repeat.

The goal is to test your way out of testing, and get to where you can scale the hell out of something.

THE CEILING

Then you hit a ceiling

As you scale and time passes, you usually hit a ceiling: a hard limit you cannot spend above efficiently.

Shape only. Find your ceiling, then push spend right back down to it.

Every dollar spent above the ceiling raises your cost per result and lowers result volume. So decrease the spend, push it right back down to the ceiling, and turn it into what we call a foundational campaign (or foundational ad set, if that is the level you run at).

The ceiling becomes the key.

TRY IT

Then you duplicate the ad set

You hit the ceiling, scaled the ad set, and have your one to three winners in it. Now duplicate it: a new ad set with the same targeting and, at the ad level, those 25 to 30 ads again. It is fine for it to sit in the same campaign, back on whatever your test budget is.

30Kept, no reach

0Pulled out

0Added new

30Total

Thirty ads loaded on a test budget. Nothing has run yet.

The ads that got no reach are not wasted. They are exactly what carries into the next duplication.

THE OTHER BRANCH

When the winners bring you garbage

This branch matters more than the first. Say the one to three winners are technically working, at a cost per result that might even look good. Then you listen to the calls, see who shows up to your webinars, or your sales team tells you these people suck.

If you do not like who is coming in, duplicate again and remove the one to three that were getting the reach. Even though they technically won, call them losers and stay away from that type of messaging. Then kill the ad set you duplicated from, because it has losers running in it and you do not want budget going to losers. Add one to three new ads, get back to seeing some victory, and repeat.

An ad that hits your cost per result and fills your calendar with the wrong people did not win. It just looked like it.

BE READY

Always have ads on the side

Keep ads sitting on the side, ready, for two reasons:

Ad fatigue hits

Relaunch the ad set with fresh ads and push past it.

An ad set has losers

When it needs duplicating, you have ads lying around instead of starting a film day from zero.

The speed at which some people film and edit new ads is intolerable. The agencies that never run into this are not faster; they stay on their clients about filming, stay on them about getting the footage over, and then edit the videos for them. Nobody in that loop gets to be the bottleneck.

And even if you only take 24 to 48 hours to turn ads around, why justify wasting 24 to 48 hours?

Ads on the side are not wasted effort. They stop you losing 24 to 48 hours the day you need one.

Testing Your Way To Greatness

The Structure