What Changed in Meta Advertising?

Andromeda Explained for Advertisers (2025 Guide)

Short answer

Meta fully changed how ads are selected for delivery.

Audience targeting is no longer the main optimization layer.

Creative-to-audience matching is.

This change is powered by Meta’s internal system called Andromeda.

What Is Andromeda in Meta Ads?

Andromeda is Meta’s core ad delivery infrastructure, not a campaign type.

It is a deep neural network that:

• Analyzes how users interact with ads

• Decomposes creatives into behavioral signals

• Matches creative attributes to users automatically

In practical terms, Andromeda decides:

Which creative should be shown to which user — before classical metrics like CPA or ROAS exist.

When Did Meta Roll Out Andromeda?

Andromeda rollout timeline:

• Spring 2024 — Private beta with large advertisers

• December 2024 — Quiet public rollout

• July 2025 — Most accounts experienced delivery changes

• October 2025 — Global rollout completed

Meta never announced Andromeda publicly. The system was introduced gradually inside the existing ad infrastructure.

How Meta Advertising Worked Before Andromeda

Before Andromeda, Meta ads were audience-first.

Old optimization model:

1. Advertiser selected audiences (interests, Lookalikes)

2. Machine learning expanded reach inside those audiences

3. Creative was optimized after delivery started

Typical setup before:

• Many campaigns

• Many ad sets

• Hyper-segmented audiences

• Lookalike audiences (1% being most effective)

• Up to 50 ads per ad set

This model relied on who you targeted first.

What Changed With Andromeda?

Core change: priority inversion.

Before	Now
Audience → Creative → Results	Creative → Audience → Results
Targeting skills dominant	Creative diversity dominant
CPA / ROAS optimization	Delivery & interaction optimization

Andromeda evaluates creative signals first, then finds users likely to respond to those signals.

Why Many Meta Ads Now Get Zero Delivery

This is one of the most common Andromeda effects.

Reason: creative similarity suppression

When multiple ads share similar signals (visuals, pacing, messaging), Meta:

• Selects one ad as a representative

• Concentrates spend on it

• Suppresses similar ads

  
  

How to identify this in Ads Manager:

• One ad receives most of the spend

• Other ads receive almost no spend

• Meta suggests “Related Media” inside the winning ad

This means:

Meta considers those ads functionally identical.

What Meta Now Considers “Different” Creatives

Andromeda does not care about:

• New copy lines

• New fonts

• New backgrounds

• Slight hook variations

These are not different creatives to the system.

What actually counts as creative diversity (angles):

An angle is a distinct creative approach based on meaning, not appearance.

Valid angle changes include:

• Pain-focused vs outcome-focused (“Point B”)

• Educational vs emotional framing

• POV UGC vs scripted narrative

• Founder-driven vs customer-driven

• Objection-handling vs aspiration-driven

Different angles = new signals for the algorithm.

How Creative Strategy Changed Under Andromeda

Old Creative Process

• Find one winner

• Iterate variations

• Optimize micro-details

New Creative Process

• Produce multiple distinct concepts

• Let Meta learn how each concept maps to users

• Scale diversity, not repetition

You can now upload up to 10 different creatives inside a single ad — Meta evaluates them independently.

How Meta Matches Creative to Audience Now

Andromeda analyzes:

• Visual style

• Hook behavior (first seconds / first frames)

• Emotional trigger

• Message intent

• Interaction probability

It then matches those signals with:

• User behavior patterns

• Engagement history

• Contextual likelihood to act

This matching happens before CPA, ROAS, or CTR stabilize.

Why CPA and ROAS Became Lagging Indicators

CPA and ROAS:

• Reflect what already delivered

• Do not explain why delivery happened

• Appear after optimization decisions

Under Andromeda, advertisers need leading indicators:

• Is the creative unlockable?

• Is Meta expanding reach?

• Is delivery collapsing into one ad?

This is why creative diagnostics matter more than pure performance metrics.

How Winning Teams Now Produce Meta Creatives

Effective creative production systems include:

Idea Generation

• Scraping competitor ads at scale (Ads Library + AI)

• Customer development interviews

• AI transcription and analysis of sales calls

• Mining objections and phrasing used by real users

Filtering

• Remove ideas that duplicate existing angles

• Prioritize emotional and structural contrast

Production

• Change structure, pacing, format — not just copy

• Treat each creative as its own hypothesis

Talent Pool

• Multiple actors / faces matter

• UGC diversity creates signal diversity

• Agencies with creator rosters scale faster

Non-Obvious Meta Ads Changes Under Andromeda

• Retargeting is no longer always superior

• Manual placement selection rarely helps

• Fewer campaigns + fewer ad sets perform better

• Data density inside an ad set matters more than structure complexity

  
  

Key Takeaway

Meta ads did not stop working.

The optimization layer changed.

• Audience targeting became background logic

• Creative became the primary input

• Diversity replaced iteration

• Production systems replaced guesswork