What is Meta Andromeda and how does it affect arbitration?

Hello, community! Meta Andromeda is no longer some shiny innovation. It started creating problems for affiliates back in April 2024. Enough time has now passed to assess its impact on ad campaigns more or less objectively. 

If you actively run Facebook ads, you already know what Meta Andromeda is. For the rest — our condolences here’s a quick explainer.  

And a reminder: we recently published a guide on how to create and effectively grow a Facebook Business Page in 2025. Highly recommended! 

What is Meta Andromeda? 

Meta Andromeda is a machine-learning stack — essentially a neural network — that automatically optimizes ad campaigns. The issue is that its optimization sometimes runs counter to the affiliate’s intent. Even creatives that convert well can suddenly stop serving, replaced by different ads that Meta Andromeda “believes” fit the audience better. The audience — not you. 

As soon as the algorithm rolled out in April 2024, many “traditional” connections simply stopped working, causing affiliates to take losses. Everyone had to test new hypotheses because nobody really understood how the model worked, what it prioritized, or how to configure ads so they’d actually reach the target audience. In other words, no one understood how the “new” ads worked. 

This was despite Meta positioning the tool as one that would better tailor campaigns to a user’s interests via personalization. For example, if a user is primarily interested in buying a car, the system would also show adjacent products like liability insurance or winter tires. 

In practice, it didn’t work that way. Reaching traffic got noticeably harder. Campaigns behaved strangely: one day a given ad wouldn’t serve at all, and the next day it would burn the entire spend. There were no clear explanations why. 

So, some time has passed. Has anything changed — is Meta Andromeda working as intended now? The answer is no; the problems are still significant. 

What problems does Meta Andromeda cause today? 

Meta assured everyone that the neural network’s core goal was to automate campaign setup. Allegedly, an affiliate could just stack creatives into an ad set and the algorithm would test creatives, find the most viable connections, and drive profit. That’s not how it turned out.  

Instead, affiliates have to create even more ad sets than before, loading each with one or two creatives. Only then can you reach the needed target. There’s zero hope the algorithm will “understand” your campaign goals on its own. In short, you end up doing the very work Meta Andromeda was supposed to eliminate. Paradoxical. 

It’s also important to note that solo affiliates with small budgets suffered the most. Many experienced buyers emphasize the algorithm isn’t hopeless: after roughly a week and about a hundred conversions, it can optimize a campaign toward profit. But that implies you’ll spend $350–$400 just on tests and optimization if your daily spend is $50 per account. For a beginner, that’s a lot. 

This is why Meta Andromeda filters out newcomers, even though automation was supposed to help them — less manual work, the neural net does the rest. But no.  

Did affiliates get any benefits from Meta Andromeda? 

Is it really all bad — did machine learning bring no upside? Early on, sentiment was very negative, but over time some potential emerged. Specifically: 

1. Meta Andromeda can sometimes discover non-obvious audiences — segments you didn’t initially target but that end up highly profitable. However, as noted above, that takes budget. You’ll have to start broad to let the system find those hidden pockets;
2. The algorithm performs somewhat better with larger budgets. If it collects enough pixel data and you feed it high-quality (read: expensive) creatives, you can realistically increase sales;
3. There’s a stronger emphasis on creatives. Maybe not a pure “benefit,” but if you’ve found a truly strong creative with high conversion, it’s a bit easier to scale — which is a real plus.

Even so, the advantages don’t cancel out the drawbacks, namely:  

1. Risks tied to the algorithm’s opacity. Affiliates simply don’t know how ads truly work now — and we’re talking about one of the most expensive channels in the market;
2. Small budgets are left out in the cold. They weren’t invited to the party;
3. Even big campaigns feel the pain, since traffic gets pricier as competition rises. To beat a higher-quality creative, you’ll overpay at the auction;
4. Lack of predictability — you can’t forecast campaign outcomes.

Of course, you can adapt — even to the changes Andromeda brought. Right now, those who managed to adjust the fastest are the ones seeing the most profit.  

But it’s also clear they aren’t sharing their secrets yet. So there’s still no public, concrete playbook for operating under Andromeda with small budgets. For now, it’s watch and experiment.  

Conclusion 

Affiliates greeted Meta Andromeda very negatively — for good reason: opaque algorithms no one understood and connections that suddenly stopped working. That led to losses with no obvious way to recover, and the picture didn’t improve much: no transparency, no planning, no predictability. 

Gradually, people got used to Meta Andromeda. But did its downsides disappear? Of course not. So we’re left waiting for the algorithm to work better with small budgets. Teams can still extract quality traffic, but solo beginners remain sidelined. Who knows when that changes? 

What’s your take on Meta Andromeda? Share your experience in our Telegram community — the best affiliate community in Ukraine! 

With respect, Your Geek! 

What is Meta Andromeda — FAQs