Hey! Pep here.
Did your Facebook ads get a little weird around August—maybe CPMs went down, or new-customer acquisition tanked?
Sounds like you’ve met Andromeda.
Not many people are talking about it, which is no surprise because Meta’s biggest backend shift since iOS 14 didn’t get a splashy press release. Instead, they quietly rebuilt the engine while we (the media buyers) were driving.
So in this post I’m breaking down everything you need to know, including:
- What changed: How Andromeda is different from past algorithms
- What’s broken: Why most of your current Facebook ad strategies are now obsolete
- And what to do now: How to restructure your account to see better results
Let’s dive in.
What actually changed (in plain English).
Andromeda is Meta’s new retrieval system—the step that shortlists which ads even get considered for the auction.
Before: the system pulled from a few hundred ads.
Now: it can scan thousands or even tens of thousands of ads in milliseconds, using sequence learning (what you did yesterday/last week) to decide what you’ll likely click today.
Think of Andromeda as the engine, and Advantage+ as the cockpit interface that feeds it broad signals and creative density.
In other words, Meta wants:
- Fewer knobs, more inputs
- Broad reach with automated placements/budgets
- And a bigger, more diverse creative library
The new rules of the game.
1) Creative portfolios > single “winner”
In this new system it’s no longer recommended to chase after the big winner. Now, creative variety is the #1 lever.
So instead of split-testing the “Buy Now” or “Learn More” button…
We’re now rewarded by having a variety of concepts and formats (problem/solution, founder stories, UGC, etc.) and targeting a variety of customer personas.
It’s not enough to just do headline variations; you need to start loading different angles.
2) Consolidate campaigns for data density
Forget everything you know about campaign structure.
We used to only want 3–6 ads in an ad set, but now that Meta is rewarding creative portfolios, we’re seeing campaigns with more creatives (even as many as 50 ads in a single campaign) absolutely crushing it.
The new formula is simple:
- One campaign per objective (sales, lead gen, etc.) per frontend offer
- Broad targeting with Advantage+ defaults
- 10–50 distinct ads in a campaign (and added to the same campaign)
The big takeaway? Consolidation is king.
In order for Andromeda to succeed it needs data density, so you don’t want to deprive the algorithm of information by splitting your budget across 10 campaigns.
3) Become an “input curator,” not a micromanager
The machine is truly starting to take control here, guys.
If we give the algorithm enough creatives and data to work with, AI will handle budgeting and audience segmentation.
Our job is to:
- Feed new creatives weekly
- Prune true losers
- And quit micro-managing
Meta says to test 20 creatives/week. These can go straight into your evergreen campaign, or you can put new creatives in a testing campaign and then move winners to your evergreen.
Healthy campaigns show spend distributed across multiple ads.
Oh, and don’t smash the kill switch after 24 hours—give it 7+ days unless the campaign as a whole is missing benchmarks and Meta has already starved the ad.
What’s working right now (from live accounts)
One of the huge benefits of running our agency is we don’t have to rely on theory; we get to see what’s working (and what’s not) in real-life ad accounts.
Here are 2 client case studies on the effects of Andromeda.
Case study #1: Consolidate and reload the library
One of our clients, Clean-eez, saw their performance tank in mid-August, with an increase in CPMs, CPCs and overall CPAs.
Here’s what their ad account looked like at the time:
So we consolidated their ad account from 8 campaigns down to just 2 campaigns (1 per frontend offer).
We took every strong ad from the past year—and even some they turned off due to fatigue—and loaded them into 1 new campaign with creative diversity: founder story, behind the scenes content, customer testimonials, everything.
Here were the results: CPMs -20% and CPAs -35% within days, with spend spread across many creatives instead of one hero hogging 90%.
Case study #2: Beware the “phantom” new-customer growth
Another client account looked great in Ads Manager, with strong CPAs and ROAS and an increase in repeat customers.
But here was the problem: Shopify said new customers were down.
It turned out, even with broad targeting Andromeda was hyper-optimizing for the easiest conversions (site visitors and previous buyers). We manually excluded these segments and new-customer volume rebounded immediately.
The big takeaway: pay attention to your backend (NC ROAS, first-order profit) to ensure your ads are delivering what’s best for you, not what’s best for Meta.
Your 4 action items for Andromeda.
I’m sure this article is a lot to take in, but if you follow these 4 action items, your ad account should be in good shape.
1) Consolidate your campaigns
Use CBO with broad targeting, and a diverse creative portfolio full of different ad formats and angles.
2) Reuse old creatives
Go through your archives for creatives that performed well, but were maybe turned off due to fatigue. Add them into 1 big CBO campaign and see if they start working again. (We’ve seen this work for several clients.)
3) Double down on creative production
Creative strategy has never been more important. As I said, Meta recommends testing 20 ads/week. To increase output try using AI to make creative briefs, hiring a creative strategist, or working with an agency like Smart Marketer.
4) Keep an eye on your backend
If you see return customers increasing and new customers decreasing, add exclusions (even though that’s not Meta best practices). Meta’s goal is to get a good platform ROAS, but yours is to ensure acquisition is strong.
Final thoughts.
Andromeda is Meta’s biggest shift since iOS 14—not because of a shiny new feature, but because the engine has changed.
The advertisers who adapt (consolidate, diversify, protect NC acquisition) will ride lower CPMs and steadier scale.
The ones who cling to the 2024 playbook—with micromanaged audiences, endless copy tests, and dozens of individual campaigns—will wonder what happened to their performance.
By the way: I also recorded this post as a quick YouTube video, if you learn better that way.
Thanks for reading, and see you next time!
