Andromeda
Andromeda is the name associated with Meta’s machine-learning retrieval engine that selects which ads to consider for each opportunity — the system behind the platform’s improved ability to match ads to people from a huge candidate pool.
Andromeda sits at the retrieval stage of Meta’s ad system: from an enormous pool of eligible ads, it narrows to the strongest candidates for a given person and moment before the auction ranks them. Its improvement is a large part of why broad targeting and consolidated structures now outperform manual segmentation — the system is simply better at finding the right ad from more options.
For operators the practical implication is not a setting to change but a strategy to accept: give the system more high-quality creative to retrieve from and cleaner signal to judge with, because a better retrieval engine rewards variety and punishes thin, repetitive creative pools.
Want to know what yours is?
The free audit reads your actual account and computes the numbers behind every one of these definitions — no generic benchmarks.
Run the free audit