Meta’s advertising system increasingly uses artificial intelligence to determine which ads may be most relevant to each person. For every business relying on digital marketing, understanding these changes is important, as they can influence how Meta advertising campaigns are structured, targeted, and optimised. One of the major developments in Meta’s advertising technology is Andromeda, a machine-learning system designed to improve the retrieval stage of Meta’s ad recommendation system.
In this blog, TLM Studios Private Limited, a trusted digital marketing agency in Kerala, discusses what the Andromeda update is, how it works, and what it means for advertisers.
What is the Andromeda update?
The Andromeda system is a machine-learning system developed by Meta to improve how its advertising platform identifies and retrieves relevant ads for users. It operates within Meta’s ad recommendation system, helping to select a smaller group of potentially relevant ads from a much larger pool before further ranking determines which ads are shown.
Unlike a conventional advertising update that introduces a new campaign setting or targeting option, Andromeda is a technical system that works behind the scenes. It is designed to handle a large and growing volume of ads and creative variations on Meta’s platforms while supporting more personalised ad retrieval.
For advertisers, understanding Andromeda is important because it reflects Meta’s increasing use of machine learning and automated systems to determine which ads may be relevant to different users.
How Andromeda Works?
Massive Scale
Meta’s advertising system can have tens of millions of potential ad candidates. The retrieval stage is responsible for narrowing this huge pool down to a much smaller group of relevant candidates before later ranking systems determine which ads are ultimately shown. Andromeda was developed to handle this scale more efficiently while supporting greater personalisation.
First-Stage Filtering
Andromeda operates at the retrieval stage, which is the first step in Meta’s multi-stage ad recommendation process. It identifies a smaller set of potentially relevant ads from the much larger candidate pool. More sophisticated ranking models then evaluate these candidates to determine the final set of ads to be shown to the person.
Behavioral Focus
Meta’s advertising recommendation systems use machine learning to understand signals related to people and their interactions with ads. Meta has also described the use of event-based and sequence-based learning to better understand behavioural patterns and predict which ads may be relevant. These approaches allow the system to consider patterns of engagement rather than relying solely on manually defined characteristics.
Impact on Advertisers
Targeting is Automated
Meta’s advertising systems increasingly use automation to manage different aspects of ad delivery. Advantage+ tools can automate areas such as audience selection, budget allocation, placements, and other campaign decisions. This allows advertisers to provide relevant signals and creative assets while Meta’s machine-learning systems help determine which opportunities are most relevant for reaching potential customers.
Creative Diversity Wins
With a large and growing number of creative assets entering Meta’s advertising system, having different creative approaches can give the system more options when matching ads with different people and situations. Meta specifically designed Andromeda to handle the increasing volume of ad creatives associated with tools such as Advantage+ creative and generative AI.
This makes creative variety an important consideration for advertisers. Different messages, formats, visuals, and value propositions can provide the system with more options to evaluate rather than relying on only a small number of similar ads.
Consolidated Campaigns
Meta has increasingly encouraged advertisers to simplify campaign structures and allow its automated systems more room to optimise delivery. Its performance-marketing guidance recommends combining ad sets where appropriate and minimising unnecessary changes during the learning phase so that the delivery system has more opportunity to learn and optimise.
For advertisers, this can mean shifting the focus away from highly fragmented campaign structures and toward providing strong creative assets, clear objectives, and sufficient signals for Meta’s automated systems to work with.