Meta Discloses 'Andromeda' Ad Retrieval

As Meta continues to build out its Advantage+ suite of AI-powered ad tools, the tech giant wants advertisers to understand the inner workings of its processes and AI use within campaigns -- specifically its “Andromeda” ad retrieval system, which it says marks a significant upgrade from its traditional ad-matching engine.

Ad retrieval is the first step in Meta's multi-stage recommendation system. Using a combination of new hardware (NVIDIA and Meta Training and Inference Accelerator chips) and machine learning, Meta designed Andromeda to improve ad personalization and targeting within this initial phase by allowing the system to assess more factors when deciding which ads to show specific users across its family of social media apps. 

“This stage is tasked with selecting ads from tens of millions of ad candidates into a few thousand relevant ad candidates,” the company states in a recent announcement, adding that the following stage involves larger and more sophisticated ranking models that predict people and advertiser value to determine the final set of ads to shown to the user.