Facebook ad Experts @ Digital Sales…Andromeda: Decoding Meta’s New Ads Algorithm

At Digital Sales we like to think we are Facebook at Experts…

So, today we are going to discuss a subject that would only be known to serious Facebook Ad Experts, Facebook ads consultant, and Facebook Agencies on top of their game…Andromeda.  So, what is Andromeda, how does it impact Facebook Ad Campaigns, how to adapt and influence in order to deliver better Facebook Ad results.

Andromeda: Decoding Meta’s New Ads Algorithm

The Meta advertising ecosystem — across Facebook, Instagram and Messenger — is undergoing its biggest transformation since the introduction of machine learning.

This shift, internally known as Andromeda, marks a fundamental evolution of the ad-delivery pipeline: from a ranking-first model to a high-performance pre-ranking and filtering system.

This isn’t a small update. It’s a full re-engineering process that forces advertisers to rethink their strategy from audience optimisation to signal optimisation.

1. The AI Leap: From Ranking to High-Speed Retrieval

To understand Andromeda, you first need a clear picture of how large-scale recommendation systems work.

The Old Model (Ranking-Based)

Previously, the algorithm focused primarily on the ranking stage.  Meta would take thousands of eligible ads for a user and calculate the Estimated Total Value for each one. The ad with the highest score would win the auction.  Manual audience targeting was the starting point.

The Andromeda Model (Retrieval + Filtering)

Andromeda acts as an advanced Retrieval engine — essentially an ultra-strict gatekeeper before ranking takes place.

Key characteristics:

• Purpose: Instead of ranking thousands of ads, Andromeda must instantly select the 10–20k most relevant ads within microseconds.
• Hashing & Vectorisation: Meta is now using advanced hashing and vector-based models to convert creatives and user profiles into compact mathematical representations known as embeddings.

The system measures vector similarity between the user profile and the ad creative.  Low similarity = instant removal before the auction ever happens.

• Implication: More efficiency, but also more punishment for weak or irrelevant creatives.

2. The Creative as a Data Vector (Creative Embedding)

Traditional manual targeting (e.g., “Digital Marketing”, “Travel”, “Fitness”) has lost power because these labels are too generic for Andromeda’s vector-based understanding.

What Andromeda truly sees is the Creative Embedding — a data vector created by AI that encapsulates the essence of your ad.

What influences your Creative Embedding?

• Visual & Textual Content: What the AI identifies in your images/videos (objects, emotions, context) and what your copy communicates.
• Post-Impression Signals: How early users interact with your ad — watch time, pauses, shares, comments, etc.
• Destination Quality: Alignment between ad promise and landing page, analysed via scraping and server-side data (including CAPI).

The Verdict

Your creative, not your chosen audience, generates the buying-intent vector that Andromeda uses to match your ad to the right users.

3. The New Strategy: Optimising Your Creative Library

To win under Andromeda, your creatives must generate strong, diverse embeddings.  This demands a complete mindset shift.

Old Strategy (Ranking-Based)

• Micro-targeted audiences of 1–5M users
• A/B testing small text variations
• Aim: increase eCPM mainly through manual bidding

New Strategy (Andromeda Retrieval-Based)

• Broad targeting of 50M+ users or full automation (Advantage+)
• A/B/C/D testing of distinct creative angles, not tiny copy tweaks
• Aim: increase eCPM through higher CTR, proving your creative survived the filtering stage

The “Creative Angle Diversity” Framework

To help Andromeda find your ad across multiple user niches, you need diverse creative vectors.

Example: Accounting Software

• Vector 1 — Emotional Angle: Focus on time freedom and reduced stress
• Vector 2 — Rational/Technical Angle: Highlight compliance, accuracy and data security
• Vector 3 — Social Proof Angle: Customer testimonial showing real outcomes

Each angle generates a different embedding, increasing the chance that one will match a relevant user profile within Meta’s massive data pool.

Conclusion: Trust the System — Improve Your Inputs

Meta’s Andromeda represents a new level of AI maturity in digital advertising.  The platform no longer wants advertisers to guess their audience. Instead, it expects brands to deliver high-quality signals:

• Strong creatives
• Diverse angles
• Clear conversion intent
• Consistent landing experiences

The era of “set it and forget it” targeting is over. The new era is continuous creative optimisation, feeding Meta’s retrieval engine with the creative fuel it needs to find your highest-value customers.

Are you a Facebook Ads Expert?

Well, if you employ the above advice and optimisation steps you are on the right path to be considered a Face Ads Expert.

If you are a company seeking help with your Facebook campaigns, reach out to the Facebook Ad experts at Digital Sales and we will audit your account and explain what might be required to optimise your Facebook Ads campaigns.  See all our contact details here.