Andromeda Changed the Kill Rules: Why Cutting Your Low-ROAS Ads Is Now Killing Your Scale

The first thing I saw when I opened the account was a kill list. The previous setup had paused everything that wasn't pulling a 7x, leaving four ads standing out of what must have been forty. Clean. Disciplined. Exactly what you'd have done in 2022.

And the account was shrinking. Spend was stuck, revenue was sliding, and the blended ROAS, the number they'd been so proud of, had started to slip the week after the cull rather than hold.

I've now seen this same pattern in enough accounts to call it what it is. The discipline that built great accounts a couple of years ago is now quietly strangling them. Not because the founder got worse. Because the rules underneath changed and the playbook didn't.

The rule that used to be gospel

For years, the move was simple and it worked. Find your winners, pile budget into them, cut everything that wasn't clearing a high bar. If an ad was doing 2x and your target was higher, you killed it without a second thought. Concentrate spend on the 7x, starve the rest. More efficiency, less waste.

I'm not going to pretend that was wrong at the time. It was right. On old Facebook, the algorithm largely served the creatives you fed it to the audiences you picked, and pruning the low performers genuinely did tighten the account.

Then mid-2025 happened, and the Andromeda update changed how the whole machine decides who sees what. That one shift is why the old kill rule has flipped from helpful to harmful.

What actually happens three days after you cut

Here's the pattern I want every founder to recognise, because it's the tell.

You do the cull. You pause everything except your highest-ROAS ads. And for about two or three days, it looks like genius. Your blended ROAS jumps. The account looks cleaner and more profitable than it has in months. You feel like you finally got disciplined.

Then it tanks. Around day four or five, the whole thing rolls over and your efficiency ends up worse than before you touched it.

What happened? Those "inefficient" ads you killed weren't just weak performers sitting next to your winners. They were feeding your winners. They were doing the top-of-funnel work, putting your brand in front of people who'd never heard of you, warming an audience that your 7x ad then quietly closed. Cut the feeders and the 7x has nothing left to convert. It was never a standalone hero. It was the last touch on work the others were doing.

That's the trap. The damage shows up on a delay, so the cut and the collapse don't look connected. They are.

Stop reading ads. Start reading concepts.

The shift that fixes this is learning to read your account at the concept level, not the individual-ad level.

Quick definition, because the word gets thrown around loosely. A concept isn't just "a new ad idea" or a fresh vibe. Think of it as an angle plus a format that, together, attract a particular slice of the market. The splashy product-on-a-clean-background concept pulls one kind of buyer. The raw, hand-held, authentic-looking concept pulls another. Same product, different doors into it.

Under Andromeda, the platform is essentially matching each concept to the audience it suits. So different concepts don't just perform differently, they reach genuinely different people. That changes everything about how you judge them.

Three concepts, three jobs

Let me make this concrete with numbers. Say you've got three concepts running, and your profitability target is a 2x.

• Concept A does a 2x. On the old rule, this is borderline. But it can absorb real budget, maybe ~$1,500/day, because it's hitting the broad top of the funnel where the volume lives.
• Concept B does a 3x. Solid, mid-funnel, and it'll take maybe ~$500/day before it starts to strain.
• Concept C does a 7x. Beautiful number. But it only reaches a small, warm pocket of the market, so it'll only spend maybe ~$100/day before the well runs dry.

The old instinct screams: kill A and B, put everything into C. So you do. And now your account can spend about $100 a day, because that 7x concept was never able to hold more than that. You didn't get more efficient. You amputated 95% of your scale and kept the prettiest number on the spreadsheet.

The new way to read it: all three are above your 2x target, so all three are making you money. Stacked together they're a far bigger, far more profitable account than concept C could ever be alone. The 2x isn't your weak link. It's your engine. It's the only reason the 7x has warm people to convert in the first place.

I think this is the single hardest mental shift for founders who came up in the old era, because every instinct you trained says the 7x is the one to chase. In reality, the 7x is riding on the back of the 2x.

So what do you actually cut?

None of this means you stop cutting. It means you change what "underperforming" means.

The new bar isn't "is this below my best ad?" It's "is this above my profitability target, and is it pulling its weight at the funnel stage it sits in?" Here's how I'd make the call:

Keep anything above your target ROAS, full stop. If a concept clears your 2x (or whatever your real break-even-plus-margin is), it stays, even if it's nowhere near your hero number. It's contributing profit and it's likely feeding the funnel.

Only cut what's genuinely below target and not earning its keep. Ads that sit under your profitability line and aren't doing obvious top-of-funnel work are fair game. That's real waste. Kill it.

Read it on a delay, not in real time. Because Andromeda spreads the impact out, don't judge a cut on the two or three days right after. Give it a week or two and look at what happened to your blended efficiency and your total spend, not just the line on the ad you kept.

Watch your attribution window. A lot of the "the 7x is amazing" illusion comes from short click windows that hand all the credit to the last touch. If you can, look at the account through an incremental lens. You'll often find your "low-ROAS" concepts are doing far more of the real new-customer work than the dashboard's default view admits.

The honest summary: your job isn't to find the one best ad and bet the account on it. It's to stack a set of concepts that each clear your target at different points in the funnel, and let them compound into something bigger than any single one of them.

A quick gut check

If your account has quietly shrunk to a handful of ads and you can't push spend without efficiency falling apart, there's a decent chance you've over-pruned. The fix usually isn't a clever new hack. It's rebuilding the top of the funnel you cut away, and giving the algorithm a wider, more competitive set of concepts to work with again.

That's a lot of what we spend our days doing inside client accounts: reading them at the concept level, working out which "inefficient" ads are actually load-bearing, and rebuilding the stack so the whole account can scale instead of just the prettiest line in it.

So before you make your next round of cuts, ask yourself one thing: are you killing your weakest ads, or are you killing the ones quietly feeding your best? If you're not sure which, that's exactly the thing worth looking at before you touch the kill switch.