What Andromeda Actually Changed
The Meta Andromeda update, which rolled out through 2024 and 2025, fundamentally changed how Meta's algorithm evaluates and distributes paid creative. The changes are documented in Meta's own developer and advertiser resources. Most brands are aware that something has shifted. Far fewer understand specifically what changed and what it means for how they need to think about creative.
Before Andromeda, Meta's distribution system evaluated a manageable pool of creative candidates per impression and selected from among them based on a combination of bid price, audience targeting, and predicted engagement rate. Creative quality mattered, but the system had limits on how many candidates it could meaningfully evaluate in real time.
Andromeda removed those limits. The new system can evaluate orders of magnitude more creative candidates per impression. In practical terms, this means that when your ad is competing for an impression, it is competing against a much larger and more diverse pool than it was before. The algorithm has more options, which means it is making more precise decisions about which creative deserves distribution.
The brands that benefit from this change are the ones with genuinely diverse, high-quality creative libraries. The brands that are being hurt by it are the ones running multiple variations of the same underlying concept.
The Creative Diversity Problem
Here is the thing most brands do not realize: what looks like creative diversity from the inside often is not.
A brand running fifteen ads might have varied the copy, changed the product shot, tested different creators, swapped the background, and adjusted the call to action. From a production and trafficking standpoint, those are fifteen different ads. From Andromeda's standpoint, they may be one concept repeated fifteen times.
The algorithm evaluates creative by concept archetype — the underlying structure of the ad, not its surface variations. A problem-solution ad with five different creators and three different copy angles is still five variations of the same problem-solution concept. Andromeda recognizes that pattern. Once it has learned what your audience does with that concept, additional variations add diminishing value.
In a national brand account analyzed against Meta Andromeda distribution signals, two hundred live ads resolved to fifteen concepts the algorithm recognized as genuinely distinct. The other one hundred and eighty-five were invisible as meaningful variety. The account was producing volume. It was not producing diversity.
What Andromeda Actually Rewards
Four things show up consistently in the current distribution data as signals Andromeda rewards.
Genuine concept diversity. Different archetypes, different hook styles, different visual languages, different intended audiences within the same product universe. Not surface variation, but conceptual variation.
Persona specificity. Andromeda rewards ads that feel like they were made for a clearly identifiable type of person. Ads that feel like they could be for anyone get distributed less effectively than ads that feel specific. This is counterintuitive for brands that have always tried to appeal to the broadest possible audience, but the data is consistent.
Platform-native execution. Content that looks like it belongs in the feed — that does not visually announce itself as an ad before the viewer has decided to engage — gets distributed more efficiently than polished production content. This is not about production quality. It is about format and aesthetic alignment with what the platform's users are accustomed to seeing.
Hook velocity. The system uses 3-second view rate as a direct quality signal. Ads that fail to create a reason to keep watching in the first three seconds get deprioritized. The benchmark range Meta publishes is 20–40% as good, and above 40% as excellent.
What This Means For Your Creative Strategy
The immediate implication is that briefing for volume is the wrong objective. If your creative program is optimizing for the number of variants it can produce, it is optimizing for the wrong thing. Andromeda does not reward volume. It rewards variety.
A creative program that produces ten genuinely distinct concepts — different archetypes, different hook styles, different personas — will outperform one that produces fifty variations of three concepts, every time.
The second implication is that understanding your current concept distribution is as important as producing new creative. Before you brief your next campaign, you should know what archetypes are already in your active library, which hook styles are over-represented, and what the gaps are. Producing more of what you already have is wasted budget.
The third implication is that the pace of iteration matters. Andromeda's evaluation is continuous. As the algorithm learns from your creative library, the marginal value of each additional variation of a known concept decreases. The brands winning on Meta right now are the ones that can produce new, genuinely different concepts faster than the algorithm learns to plateau on their existing ones.
The Algorithm Is Not Waiting
The brands that understand how Andromeda evaluates creative before they launch are not smarter. They are faster. And in a system that rewards novelty and penalizes repetition, being faster is the compounding advantage.
The update is not new. The brands that have adapted their creative strategy around it are already building the data moats that will make them harder to compete with over time. The ones still briefing for volume are working harder for worse distribution.
Understanding the system is the first step. Evaluating your creative against it before you spend is the second.