Why Meta’s Andromeda update changed creative strategy for good
If your Meta ads aren't performing the way they did two years ago and nothing much has changed in your account, this is probably why.
Meta rebuilt the part of its system that decides which ads get shown to which people. It's called Andromeda. Since it landed, creative isn't the thing you make once the strategy is done. It is the strategy.
What Andromeda actually is
Andromeda is Meta's ad retrieval system. Retrieval happens before the auction. Out of the tens of millions of ads that could be shown to one person, it decides the shortlist that gets to compete for that impression. Miss the shortlist and your bid doesn't matter. You were never in the room.
The old system built that shortlist off your targeting settings and broad interest categories. It was blunt.
Andromeda reads the creative instead. Format, hook, subject, tone. Then it predicts who's likely to respond based on how people actually behave, not the audience box you ticked. Meta announced it in December 2024 and finished the global rollout around October 2025, so by early 2026 this is simply how Meta works.
Worth sitting with the scale of it. Meta says Andromeda is roughly four times more efficient at driving performance gains than the ranking models it replaced. That's a new engine, not a patch.
Why it changed targeting more than creative
This is the bit most people skip over. Andromeda didn't just make ads perform better, it changed what does the targeting.
Back when retrieval was blunt, your audience settings did the work of finding the right person, which is why building them well was a skill worth having. Now the system reads the creative and finds that person itself, inside a much broader pool. Your audience definitions have gone from gates to suggestions.
That's the mechanical reason broad beats narrow now. It isn't a trend, and it isn't a platform preference you can sit out until it passes. The matching got better than your narrowing.
The creative rule almost nobody has caught up to
Here's the practical one, and it makes half of what brands call testing pointless.
Andromeda groups near-identical creative together. Change a headline colour, swap the CTA, trim two seconds off a video, and there's a good chance the system files it as the same ad it already knows. Not a new test. A duplicate.
So making ten small variations of one concept gives the algorithm nothing new to work with. It's one ad in ten outfits.
New now means a different angle, a different hook, a different reason to care, or a different format altogether. That's what gives the system a real new signal to route somewhere new. The bar for what counts as new creative went up. Most accounts haven't moved theirs to meet it.
To make this concrete, here are 5 ads for the same product, a vitamin C serum, that Andromeda would read as five genuinely different signals: a problem-first hook, a founder talking to camera, a real before and-after, an ingredient explainer, and the product inside a morning routine. Same serum, five different arguments, five different pockets of the audience.
So what actually counts? Three examples below.
What this means for how much creative you need
More than most founders think, though not for the reason they assume.
It isn't volume for its own sake. It's that Andromeda can route genuinely different angles to different micro-segments inside a broad audience, and it does that far more precisely than the old system could. Give it one concept and it can only serve one kind of person well. Give it several distinct angles and it can match each one to whoever responds to it.
That's also a big part of why creative fatigue arrived faster. The system evaluates creative more precisely, so it works out sooner when it's pulled everything useful out of a given ad. Sharper evaluation cuts both ways. Quicker to find a winner, quicker to burn through it.
What this means for account structure
Mostly your structure needs to get out of Andromeda's way.
Broad audiences rather than narrow. Clean, fast conversion signals so the model has something accurate to learn from. Enough budget and enough genuinely varied creative that it has real choices to route between. None of it is exotic, but it's a different job than the one most founders were trained to do, which was picking the right audience.
What Andromeda is not
Worth being clear here, because this update gets oversold in both directions.
It isn't magic and it isn't omniscient. It won't fix a weak offer, rescue broken tracking, or make up for a budget too small to learn from. The mistake I'm seeing most is founders flooding the account with automated creative, switching on automation, and assuming the system will sort it out. It can't, not if the inputs are poor. Andromeda is only as good as the creative and the conversion signal you feed it.
It also didn't repeal the fundamentals. Where your account lands still comes down to your margin, your offer, and how good the product actually is. That was true before Andromeda and it's true now. What changed is the lever inside the ad account, not the business underneath it.
Where founders are still behind
Most accounts I audit are still built for the old retrieval system. Narrow interest stacks. A handful of creative concepts stretched over months. Ten near-identical variations mistaken for testing. Targeting decisions made once and never revisited.
None of that is wrong exactly. It's built for a system Meta doesn't run anymore. The brands that adjusted treat creative as the main lever and give the algorithm broad, clean inputs and genuinely varied angles. The ones that haven't are usually the ones telling me their ads stopped working, when what actually happened is the platform moved and their account stayed put.
What to do about it now
Nothing complicated. It's a reordered set of priorities more than anything.
Widen your targeting before you narrow it further. Treat creative testing as the ongoing job rather than the occasional fix, and make sure a test is a real new angle and not a recolour. Get your conversion tracking clean, because Andromeda is only as good as the signal you give it.
If your creative strategy hasn't meaningfully changed since 2023, this is the update that says it should.
Book a discovery call with me here. I'll look at your ad account and your business numbers together, and tell you straight where the real opportunity sits.
Additional resources