What the Andromeda update actually is

Everyone in the Facebook and Instagram ads space is talking about Meta Andromeda update right now. There is a lot of fear-mongering and misinformation, so it is important to understand what really changed, what impact it has on ad performance, and what you need to update in your current ad setup so you do not get left behind.

Meta first published the Andromeda article on December 2, 2024. That was almost a year ago at the time of this recording. It has only recently become a big topic because Meta rolled it out gradually. Some advertisers got it earlier, others later. It also reached some campaign types before others.

Now we are at the point where the vast majority of advertisers are operating inside the Andromeda system. Those who have not updated their setup have often seen results drop and are understandably concerned. That is where a lot of the noise comes from.

How Meta describes Andromeda

Meta explains that AI plays an important role in its advertising system. It uses machine learning to predict which ads a person will find most interesting. That sounds like standard ad platform language, but it is key to understanding Andromeda.

Andromeda is essentially a new, more sophisticated ads recommendation system. It changes how ads are retrieved and ranked, and that has consequences for how we as advertisers should structure our campaigns.

Meta describes this in stages:

1. Retrieval
   Retrieval is the first step. The system selects a few thousand relevant ad candidates from tens of millions of possible ads.
2. Ranking
   Larger and more sophisticated ranking models then predict both people value and advertiser value to decide which specific ads to show each person.

If you target a broad audience, like the entire United States, you are competing with huge numbers of advertisers. Andromeda’s job is to cut that down from millions of candidates to the handful that are most likely to be relevant and effective for each individual user.

Why Andromeda exists: dealing with AI creative and latency

Meta highlights two big challenges in the new AI-heavy advertising era:

1. Explosion in ad variations
   Advertisers now use AI tools both outside and inside Meta to create more creative variations than ever.
   Inside Meta, features like Advantage+ Creative automatically generate multiple ad variants by:
   • Adding text overlays
   • Animating images
   • Adding music
   • Applying various enhancements
   A single “seed” ad can turn into many different ad variations. Add to that the creatives produced by external AI tools, and the number of ad candidates explodes from millions to hundreds of millions, even billions. Andromeda is designed to handle and use that explosion, not be overwhelmed by it.
2. Tight latency requirements
   Meta has to make all these decisions incredibly fast. When someone is scrolling or swiping through Stories, the next ad has to load instantly.
   The system must:
   • Filter enormous numbers of ad candidates
   • Pick the right one in real time
   • Take into account changing user interests and context

Andromeda uses deep neural networks and high-performance hardware (like Meta’s custom Grace Hopper chips) to do this at speed. The technical details are complex and more relevant to engineers. For us as advertisers, the key is the outcome: a system that delivers more personalized ads and aims to maximize return on ad spend (ROAS).

Think of Andromeda like Netflix recommendations

A helpful way to think about Andromeda is to compare it to Netflix.

Netflix does not show the same comedy shows or films to everyone. It recommends different content first to different users based on their viewing history and what they are likely to enjoy.

Before Andromeda, Meta’s system already tried to match ads to users, but it did so more at a group level. People were bucketed into interest groups (for example, interested in A, B, and C), and ads were shown based on those group classifications.

With Andromeda, the system gets much more individual:

• It personalizes at the person level, not just the audience segment.
• It behaves more like a streaming recommendation engine than a simple audience filter.

The goal is to understand which specific ad each individual user is most likely to find interesting and which ad is most likely to help the advertiser hit their objective (leads, sales, etc.).

Meta’s incentives: why Andromeda is not here to “hurt” advertisers

There is a myth that Meta built Andromeda to worsen results and squeeze advertisers. That does not make sense logically.

Meta is highly incentivized to help advertisers get good results. We are their customers. If we stop seeing leads and sales from Meta ads, we stop spending. If that happened on a large scale, Meta’s ad revenue would drop and the company’s value would fall.

So the real reasons for Andromeda are:

• More personalization
  Show people more relevant ads, which they are more likely to engage with and which deliver better results for advertisers.
• Better timing
  Serve the right ad to the right person at the right moment. Timing matters. In many industries, even if a person is a perfect fit, if the ad arrives after they have already purchased elsewhere, the opportunity is gone.
• Necessary system upgrade
  With AI producing far more creatives, Meta needed a more powerful recommendation system to handle the volume and turn it into better performance, not chaos.

“If Andromeda is so good, why are my results worse?”

Not everyone has worse results. Many advertisers are seeing similar or even better performance than before. But some have definitely seen a drop.

There are a few key factors to consider:

1. Seasonality
   Results may drop because of timing, not Andromeda. For example:
   • Q4 CPMs are typically higher because of Black Friday and Christmas.
   • Your business may naturally perform better in some seasons than others.
2. Market changes
   Competition, consumer behavior and pricing pressures can all change independently of any Meta update.
3. Your strategy has not been updated
   This is the big one you can control.
   Andromeda changes how the system works. Best practices have shifted. If you keep running your old structures and habits, you can get punished by the new environment.

The opportunity is that if you do update your strategy, you may not only recover but actually outperform your previous results. And if your competitors are slow to adapt, you gain a competitive edge.

What needs to change: how to adapt to Andromeda

The core of Andromeda is increased personalization. That means you need to give Meta’s system more high-quality options to work with.

1. Increase ad volume

Meta used to recommend no more than about six ads per ad set. Now, with Andromeda, it can handle far more. It is common to run ad sets with 20 or more ads.

The logic:

• With more ads, Meta can better match specific creatives to specific people.
• Andromeda can say, “This ad is right for this person; that other ad is better for someone else.”

However, you do not want random or low-quality volume. You want many good creatives.

2. Increase creative diversity, not just variations

It is not enough to produce 20 slightly different versions of the same idea. You want different angles, not just different backgrounds.

Think in terms of:

• Stages of awareness
  Create ads for:
  • People not yet aware of the problem
  • People aware of the problem but not the solution
  • People aware of solutions but not your product
  • People who already know you and just need a push to buy
• Different pain points
  If your offer solves multiple problems, make ads that:
  • Emphasize cost savings
  • Emphasize time savings
  • Emphasize ease of use
  • Emphasize reliability, security, convenience, etc.
• Different desires
  People buy for different reasons:
  • To save time
  • To save money
  • To feel confident
  • To feel relaxed
  • To feel exclusive or “ahead” of others
  The same product can appeal in different ways to different buyers.
• Different avatars in your audience
  Use the gym example:
  • Weight loss
  • Muscle gain
  • Sports performance
  • Injury prevention or rehab
  • Older people wanting mobility
  Each group can have its own ads with language and visuals that speak directly to that situation.

The more meaningful creative diversity you have, the better Andromeda can personalize.

3. Let Meta handle sequencing

Another piece that is not always explicitly stated in the docs, but is clearly part of the direction, is automated sequencing.

Some people are ready to buy right away. For them, a direct “bottom-of-funnel” ad (lead, sale, booking) is ideal. Others need:

• Awareness style ads first
• Then more educational or problem/solution content
• Then an offer or direct call to action

Instead of manually splitting all this into separate campaigns and complicated retargeting setups, you can:

• Put different funnel-stage creatives into the same structure
• Let Andromeda decide who sees what and when

This requires more trust in Meta’s system, especially if you come from a background of highly manual, multi-step retargeting. But many advertisers who have embraced simpler structures plus high creative diversity are reporting significantly better results.

How many ads should you run now?

A common question is: “How many ads should I have in my campaign?”

The practical answer in the Andromeda era is:

As many high-quality, meaningfully different creatives as you can realistically produce.

Important notes:

• Do not run more ads than your budget can properly test at once.
• Do not churn out low quality or trivial changes just to hit a number.
• Focus on quality and angle diversity: different hooks, messages, visuals and benefits.

Most advertisers are limited by how much good creative they can create, not by what Meta can handle technically.

Simpler campaign structures, not more complex

One mistake appearing in the Andromeda conversation is over-engineering campaign structures.

Some people are proposing extremely complex setups in an attempt to “beat” Andromeda. These setups are often impractical, especially for agencies or advertisers running many accounts.

In reality, with Andromeda you generally want simpler campaign structures:

• Fewer campaigns
• Fewer ad sets
• More creative diversity inside those ad sets

Let Meta’s system do its job. Give it variety in creative and clarity in objective, rather than trying to micromanage every little segment with dozens of tiny campaigns.

Final summary

Andromeda is Meta’s new, more advanced recommendation system for Facebook and Instagram ads. It is designed to:

• Handle a massive explosion in ad variations
• Deliver more personalized ads
• Improve ROAS for advertisers
• Respect tight latency constraints so everything still feels instant for users

If your results have dropped, it is likely because:

• You are hitting seasonal or market changes
• You have not yet updated your strategy to match the new system

To thrive under Andromeda, you should:

1. Increase the volume of ad creatives.
2. Increase creative diversity across awareness levels, pain points, desires and audience avatars.
3. Simplify your campaign structure and trust Meta more with targeting and sequencing.

If you give Andromeda what it needs, it can become an advantage rather than a problem, and you may end up with better performance than before the update.

Need help running Meta ads?

If your Meta ads aren’t performing the way they used to, or if you want to make sure you’re taking full advantage of the Andromeda update, our team can help.

Rêve Marketing Agency is an English-speaking marketing agency in France specializing in paid advertising, social media strategy, and SEO. We’ve helped brands across the French Riviera and internationally scale through data-driven creative and smart optimization.

Explore how we can help your business grow.

Contact us to discuss your Meta ads strategy today.