Large recommendation systems commonly separate retrieval from ranking. Andromeda handles retrieval, efficiently identifying potentially relevant candidates from an enormous ad pool. Ranking models then evaluate this smaller set more precisely. Together, these stages form the core AI infrastructure underlying Meta’s automated advertising delivery.
Business applications and impact#
More efficient computation allows the system to consider more candidate ads more quickly, potentially improving recommendation relevance and monetization efficiency together.
Implementation guidance and considerations#
Andromeda is internal infrastructure rather than an advertiser-controlled setting. Understanding its role helps advertisers interpret the broader shift toward automated ad delivery.