Meta Andromeda Explained: What Changed in Facebook and Instagram Ads and How to Adapt
Andromeda is Meta's ad retrieval system that decides which ads are even considered for each person. Here's what it does, what it means for targeting and creative, and how to adapt.
Table of contents
Short answer: Andromeda is Meta's machine-learning system for ad retrieval, the first step that narrows tens of millions of eligible ads to a few thousand for each person before ranking decides which one to show. Meta announced it in December 2024, reporting a 10,000x increase in model capacity, +6% recall and +8% ads quality on selected segments, and built it to power Advantage+ automation. In practice, your creative now does more of the targeting: genuinely different ads reach different people, while near-identical variations compete with each other. Adapt with simpler structures, broad or Advantage+ audiences, diverse creative concepts and accurate tracking.
This guide is part of our Facebook ads cost series. For hands-on help, see our Meta ads service.
How Meta chooses which ad to show
Meta's ad delivery works in stages:
- Retrieval: from all eligible ads, select a few thousand that might be relevant to this person right now. This is Andromeda's job.
- Ranking: score those candidates on predicted outcomes and value.
- Auction: combine bids, predicted action rates and ad quality to pick the winner.
If your ad isn't retrieved, it can't be ranked or win the auction, so retrieval matters a lot.
What Meta says Andromeda changed
From Meta's engineering announcement:
- Bigger, more personalised models: a 10,000x increase in the complexity of models used for retrieval
- Better retrieval: +6% recall and +8% ads quality improvement on selected segments
- Built for automation: designed for Advantage+, where automated targeting, budgets and creative generation greatly increase the number of eligible ads
- Efficiency: large improvements in inference speed and efficiency, using newer hardware
Meta also noted its plans to make retrieval deliver a more diverse set of ad candidates.
What it means for advertisers
1. Creative is a targeting signal. With broad targeting, Meta learns who each ad is for from the ad itself and how people respond. An ad aimed at first-time home buyers and one aimed at landlords can reach different people from the same audience settings.
2. Variety beats volume of near-duplicates. Ten versions of the same video with different text colours give the system little new information. Five genuinely different concepts (testimonial, demo, problem–solution, founder story, comparison) give it more ways to match people.
3. Narrow manual targeting matters less. Detailed interest stacking can restrict what the system can do. Broad or Advantage+ audiences let retrieval find likely buyers. See Facebook ads targeting.
4. Signals matter more. Accurate conversion tracking through the Pixel and Conversions API helps the system learn what success looks like. See Meta Pixel and Conversions API.
5. Consolidation helps. Fewer campaigns and ad sets with more budget and more diverse ads usually learn faster than many small ones. See the learning phase.
How to adapt your Meta ads
| Old approach | Andromeda-era approach |
|---|---|
| Many ad sets, each with narrow interests | One or a few ad sets with broad or Advantage+ audiences |
| Many tiny variations of one ad | Several distinct creative concepts and formats |
| Creative chosen once, run for months | Steady pipeline of new concepts, retire fatigued ones |
| Optimising for clicks | Optimising for conversions with accurate tracking |
| Targeting does the segmentation | Creative and messaging do the segmentation |
A practical creative framework
Build each new batch around different angles and formats:
- Angles: problem–solution, testimonial, demonstration, comparison, objection handling, founder or team story, offer-led
- Formats: short vertical video, static image, carousel, UGC-style creator video
- Audiences in the message: name different customer types in hooks ("For busy parents…", "If you run a café…")
Then let performance data tell you which angles to expand. See Facebook ad creative and Advantage+ campaigns.
How to brief creative for Andromeda
When creative is doing more of the targeting, the brief matters more. For each new batch, write down:
- Who the ad is for: one specific customer type, described in plain words (for example "parents buying a first bike for a child", not "families").
- What they're thinking: the problem, worry or desire that would make them stop scrolling.
- The angle: one of the angles above, chosen because it suits that person.
- The proof: a review, result, demonstration or credential that makes the claim believable.
- The format: video, static, carousel or creator-style, chosen for the message.
Aim for each new ad to differ from the others in at least two of these. A testimonial video aimed at first-time buyers and a comparison carousel aimed at upgraders are genuinely different; two testimonials with different background music are not.
How to read results in an Andromeda world
Because the system now decides much of who sees each ad, look at results per ad rather than only per ad set. Ads that win usually do so because they reached a responsive group of people, so look at the audience breakdowns (age, gender, placement, region) for your best ads to understand who they found. That tells you which customer types to brief next, and which angles to retire. Give new ads enough delivery before judging them, and avoid pausing everything that isn't the top performer, since the system benefits from a range of options.
A worked example
A furniture retailer runs twelve ad sets with stacked interests and a dozen near-identical image ads. After restructuring to one Advantage+ campaign with broad targeting and eight distinct concepts (small-space solutions, a customer home tour, a delivery-and-assembly explainer, a price comparison, and so on), it finds that different concepts reach noticeably different audiences, and cost per purchase becomes more stable. (Illustrative.)
Common mistakes
- Blaming Andromeda for problems caused by poor tracking
- Producing many near-identical variations
- Over-segmenting with narrow ad sets
- Letting creative go stale
- Judging new creative before it's had enough delivery
The bottom line
Andromeda changed how Meta finds candidate ads for each person, making creative diversity and accurate signals more important than detailed targeting. Simplify structure, go broad, feed the system genuinely different creative ideas and track conversions properly.
Want your Meta campaigns restructured for how delivery works now? Request a free growth audit.
Frequently asked questions
What is Meta Andromeda?
Andromeda is Meta's machine-learning system for ad retrieval. It selects a few thousand relevant ads from tens of millions of candidates before Meta's ranking models decide which ad to show each person.
When did Meta Andromeda launch?
Meta described Andromeda on its engineering blog in December 2024 as its next-generation personalised ads retrieval engine supporting Advantage+ automation.
How does Andromeda affect my Facebook ads?
Because the system matches ads to people based largely on the ads themselves and conversion signals, creative diversity and accurate tracking matter more, and narrow manual targeting matters less.
Do I need more ads because of Andromeda?
You need more different ideas rather than more near-identical variations. Distinct concepts, angles and formats give the system more ways to reach different people.
Is Andromeda the same as Advantage+?
No. Advantage+ is Meta's suite of automated campaign tools. Andromeda is the underlying retrieval system that helps those automated campaigns find relevant ads for each person.
Sources
- #Meta Ads
- #Facebook Ads
- #Advantage+
- #Creative Strategy
Written by
SEO, AI search, paid media and social specialists
Articles from the MangoGrow editorial team are researched and written by our specialists in SEO, AI search, paid advertising, content, and social media, and reviewed before publishing.
More from MangoGrow