What the Andromeda Algorithm Actually Rewards
If Facebook’s algorithm feels like it’s working against your account, it’s usually because you’re doing things that quietly deprioritize you. I’ve spent over $100 million in the Andromeda era, and the pattern is always the same. The brands that can’t get their ROAS to a sustainable point, or their CPA down to where it needs to be, are almost all making the identical mistakes.
The algorithm grades 2 things
Most advertisers think ROAS is all the algorithm cares about. There’s a second piece almost nobody considers.
Andromeda-era Facebook asks 2 questions about every ad.
- Can this ad hit the efficiency in this account better than the other ads in that subset?
- Can this ad actually scale?
That second question explains something you have definitely seen. An ad drives an insanely high ROAS but only picks up a few dollars in spend. The moment you force budget against it, everything tanks.
It’s also why you hit a wall when you scale. You can sustain $100 a day, $500 a day, $1,000 a day, $5,000 a day. Then you go from 1,000 to 1,200 and your ROAS gets worse than it was before. Now you’re making less money than you were at the lower budget.
Most structures are broken in one of 2 directions
Once you’re accounting for both efficiency and scalability, the structure almost everybody uses stops making sense. Brands are either oversimplified or way too complex.
The old way was ad set budget optimization. You set every ad set budget individually and keep all the control, which takes a massive amount of weight away from the algorithm. Ad sets get far more budget than they should, your highest-ROAS ad sets sit capped, and proven scalable ads get nothing.
Most people are past that. What concerns me more is the opposite extreme. One campaign, one ad set, every ad thrown in together is EVEN WORSE.
In that setup you can only do 2 things. Launch ads and pause ads.
Meanwhile a whole set of tools sit unused. ROAS goals. Target cost per result. Value rules, which literally will not work with a single ad set because there’s no differentiation. Minimums and maximums, which almost nobody sets properly.
A tiered system gives you the levers back
Campaign budget optimization sets the budget at the campaign level, which is far more algorithm-friendly. People just got carried away and collapsed everything into one ad set.
Our prospecting setup works because every new round of creatives launches into a new ad set. Nothing we’ve already launched gets interfered with, only budget distribution.
That separation is what makes the tools work. An ad set geared to a specific age group or gender can carry its own value rules. You can’t do that when everything is grouped together.
It also solves fatigue. When an ad set that’s run well for months starts slipping, you normally get an ultimatum: keep it active and lose money, or pause it and hope the budget redistributes properly. Perfect distribution very rarely happens. With separated ad sets you set minimums and maximums on the fatigued one instead.
We also launch every new ad set with a 7-day minimum budget, which forces the new creatives to take a small percentage of a test budget. One of the worst things in this business is spending real money on creative that ends up with $0 in ad spend.
That’s what a pack-based system buys you. Levers, and safety nets for the day something blows up. You run out of stock, the offer disappears, the sale ends. Every growing business hits that point, e-com or lead gen, and you need something to pull.
Volume of creative doesn’t matter. Quality at scale does.
You’ve been told creative volume is critically important. I’m going to put an end to that.
At The Moonlighters we run over 80 brands spending $30M+ a month across our portfolio, and we’ve focused entirely on quality creative, then multiplied that quality with AI.
The workflow we’ve had the most success with is built into Superscale. The prompt we use most is along the lines of: use my top 5 highest spending image ads, create iterations, use my competitors as inspiration based on what’s working for them, and produce slight variety.
It pulls all the ads, knows their performance, researches our competitors, and has our entire brand kit and colors grouped in. A recent run gave us an output of around 20 ads.
The point isn’t that you only create with AI. It’s that you multiply your high-quality winning ads with AI. That’s what gives you a high hit rate and creative volume at the same time.
Then leave it alone
Once the structure is algorithm-friendly but still gives you control, and creative is multiplying through AI and real production, your job is to allow the algorithm to run.
This is where most people make the biggest mistake. Things are working, they’re above target, they’re scaling, and then they start to tweak too much. Bid caps. Cost caps. New campaigns. New structures. It’s an itching sensation.
That is not what you need to do. Sit back and feed the inputs. The highest quality creative at scale, and the highest quality offers, landing pages, and product. A strong product people actually want, and a fair exchange for their cash.
