What is Andromeda?
At its core, Andromeda is a machine learning system designed to improve ad personalisation. It powers the ad retrieval stage of Meta’s ad delivery system – the crucial first step that determines which ads make it into the consideration pool.
How does Andromeda work?
Every day, advertisers upload millions of ads. Andromeda rapidly narrows that enormous pool down to a few thousand candidates and does it faster and more intelligently than any previous system Meta has used.
This speed and scale matter because Meta’s AI tools (Advantage+, I’m looking at you), have massively increased the number of creative variations that advertisers are able to produce. Andromeda is the engine built to handle that explosion.
Why does this matter for marketers?
Meta’s delivery engine has become faster and smarter – and that gives digital marketers a huge, unprecedented opportunity to connect the right message with the right person. But to take advantage of it, we need to feed Andromeda what it wants most.
Creative Variation
And not the kind you might be used to. What doesn’t count as meaningful variation anymore:
-
Swapping one CTA for another
-
Changing a shirt from red to blue
-
Testing ‘yoga interests’ vs ‘fitness interests’
These tweaks aren’t irrelevant, but they’re no longer the levers that move performance. The truth is, the ‘perfect’ combination of copy and creative never existed – because the winning combination for one person may be the losing combination for another.
Targeting in-the-ad
Instead of targeting parameters such as age, gender and interests, marketers need to focus on producing ads that speak to different audience needs, motivations and contexts. Andromeda will do the rest.
If you provide the system with creative diversity, it is far more likely to show someone an ad that feels relevant. If you only provide a single video, Meta can only run with one angle. If you create ads that lean into different pain points, personas and concepts, the system learns who responds to what and deliver ads that feel more relevant and personal to that user in that moment.
What counts as real creative diversification?
We’re not talking about multiple versions of the same idea, Meta’s Advantage+ tools already handle that. We’re taking about:
-
Different emotional angles
-
Different audience personas
-
Different problem-solution narratives
-
Different formats and storytelling styles
This is creative variation on a scale we haven’t had to think about before.
Key takeaways
-
Andromeda changes the game at the retrieval stage, optimising the ads that are expected to provide the best campaign performance.
-
Creative variation is now the most powerful lever – not micro-targeting or tiny creative tweaks
-
Targeting happens inside the ad, through the stories, angles and emotional drivers you choose
-
The more meaningful creative diversity you feed the system, the better it can match the right message to the right person
-
You don’t need to hunt for audiences manually – Andromeda will find them if you give it the signals it needs