Meta’s Andromeda: In Layman’s Terms
Why It Matters (For Media Buyers & E-commerce Marketers)
Andromeda has quietly changed how Meta decides who sees your ads. Here’s what matters most:
Creative now does the targeting. Small tweaks to the same ad are no longer enough.
Consolidated campaigns learn faster.
Broad beats narrow in most cases.
AI needs time and budget to optimize.
Advertisers who adjust will see steadier performance and lower long-term CPAs. Those who don’t will face volatility and rising costs.
If you’re a Meta advertiser or any kind of marketer even remotely Meta ads-adjacent, you’ve definitely heard of Andromeda. It sounds scary, but it doesn’t have to be.
Meta Andromeda is Meta's proprietary machine learning system designed for retrieval in ad recommendation, and was launched globally in July 2025. Think of it as the retrieval engine that rapidly scans through millions of ads to find which ones might be relevant to show a user.
GEM (Generative Engine Model) is another foundation model released in late 2025 that delivers increased ad performance by enhancing other ads recommendation models' ability to serve relevant ads. It's the AI brain that ranks and scores ads after Andromeda retrieves them.
The old ads delivery system didn’t have a cool name like “Andromeda,” but worked like manual audience targeting where there was a lot more levers an ad buyer could pull. You could select "Women, 25-34, Interest: Yoga" and Meta would show those ads to that preset interest group. Andromeda flipped this completely.
Andromeda analyzes the pixels in your image and the speech and visuals in your video to understand who the ad is for. Instead of an ad buyer telling Meta who to target, the creative itself becomes the targeting signal. The ad acts as a self-filtering mechanism.
This is also exactly why creative is all of a sudden exponentially more important to advertising than it once was. Meta has anecdotally said that whereas in the past, Meta ads success looked like 80% ad buying, 20% creative, it’s now flipped: 80% of Meta ads success is in the creative and only 20% is in the Meta algorithm atmosphere.
What This Means for E-commerce Advertisers
The playbook has now changed dramatically and advertisers must also adapt quickly to ensure ad account efficiency and profitability.
Creative diversification is everything: You need diverse creative assets showing different angles, narratives, and customer journey stages. In the past, a common strategy was to take a single winning image and test it with dozens of different headlines or make small iterations to the asset itself. Under the new algorithm, this tactic can actually hurt your performance and iterations must be significant enough for the algorithm to not notice it as the same as an old creative, but rather read it as an entirely net-new piece of creative that needs to find its own way as it enters the ad auction
Consolidate your campaigns: The algorithm needs more data in fewer places. Every account is different, and some see success with 50 creatives per ad set, and others with 10. Test in each individual account and do what’s best for your account, but consolidate where you can, and when you can’t, sort by something that will help the algorithm, like vague/wide audience persona, and not by something more miniscule like images in one ad set and videos in another.
Go broad on targeting: Narrow, segmented audiences actually hurt performance now because they starve the AI of data needed to learn patterns. Use interest targeting when you must, but make sure that interest audience is big enough to be able to go broader if needed, and again, make sure the audience targets actually help drive the algorithm in the direction it needs to go in
Let the AI work: The system uses sequence learning to understand where users are in their journey and what they'll do next. Try to leave ad assets on for at least 2-3 days before making a change, and only make manual adjustments if you absolutely must, but try to still let the AI do some of the work, just guide it better in the right direction.
Be patient with learning: Expect to spend more before seeing results. Small advertisers might need $300-600, larger accounts could need thousands before the algorithm optimizes properly. Find the happy medium for each account and what it needs to determine success and parse through its data. Base the minimum spend amount off of the CPA of the account.
Trusted by Agency Owners, Brand Founders, and Marketers everywhere, our advertising membership is where the top 1% of digital marketers go when they need help. Meta ads, Creative Strategy, Google Ads, CRO, Email & SMS… we’ve got experts on it all. Interested? Learn more at the button below.