A big reason is Meta Andromeda. It is not a new button inside Ads Manager. It is a behind-the-scenes system that changes how ads get selected and delivered.
Here is the simple version: the ad business on Meta is moving from “build the perfect audience” to “build the best messages and give the algorithm options.”
That shift affects everyone, from small brands to agencies to Meta itself.
So what is this ? And what does it mean ?
Andromeda sits early in the ad delivery process. Think of it like a bouncer at the door: before Meta even ranks ads to decide a winner, it first decides which ads are allowed into the “maybe” group.
If your ad does not make that short list, it never gets a real chance, no matter how great your bid or budget is.
So what changes for the ads business when that bouncer gets smarter?
Targeting loses some power
Detailed interest stacks and tiny audiences can still work sometimes, but they matter less when the system is confident it can match creatives to people on its own.
Creative becomes the main steering wheel
If you give Meta one concept, one visual style, one angle, the system has fewer ways to “find the right person at the right moment.” If you give it multiple angles, formats, and hooks, it can learn faster and route each version to different pockets of users.
Simpler structures tend to win
Splitting budgets across lots of ad sets often starves each segment of data. Consolidating can help the system learn with cleaner signals.
Measurement shifts to bigger business metrics
When delivery is more automated, micro-optimising every small number becomes less useful. Many teams move toward tracking things like cost per acquisition, margin, or overall efficiency, rather than only ad-set-level “perfect” ROAS.
The value of ad operations changes
People who used to spend hours building complex audience trees now need stronger skills in creative testing, offer strategy, landing pages, and conversion tracking.
Long story Short:
Meta is trying to make ads feel more like “recommendations” to users and more like “automatic distribution” for advertisers.
That is great when you have strong creative and clean tracking. It is painful when you do not.
Different opinions from different marketers
Some Say this is mostly good news for advertisers.
The optimistic take is that better retrieval makes ad delivery more efficient. Meta’s own engineering write-up frames Andromeda as a next-gen retrieval engine designed to improve relevance and performance at scale, powered by heavier ML infrastructure. In this view, advertisers should stop fighting the machine and focus on feeding it better ingredients: more creative variety, broader setups, and enough conversion data to learn.
Here is the most direct source explaining the system from Meta’s side: Meta Engineering: “Meta Andromeda”.
Others are saying that this reduces control and makes performance feel less predictable.
The more cautious take is that automation can create “black box” moments. When your results dip, it can be harder to pinpoint whether the issue is the creative, the conversion signal quality, landing page speed, offer mismatch, or just the system exploring. Teams in this camp still simplify and broaden, but they push harder on disciplined testing, clearer creative strategy, and patience during learning phases.
In practice, both views can be true. If your account has strong signals and a steady stream of fresh creative, the automation feels like a boost.
If you run thin budgets, have weak tracking, or repeat the same creative for weeks, Andromeda can expose those problems fast.
Here are a few patterns that line up with how Andromeda behaves:
We plan creative in batches, not one-offs.
A batch means multiple angles (price, problem, social proof, demo, comparison), multiple formats (static, short video, UGC style), and multiple opening hooks.
You are not “spamming variations.” You are giving the system real choices.
We keep the account easier to read.
Fewer campaigns and fewer ad sets makes it clearer what is actually happening, and it concentrates learning.
When something works, it scales with less internal competition.
We treat tracking like part of creative performance.
If the pixel and events are messy, the system learns the wrong thing.
Clean conversion signals are not boring admin, they are the map Andromeda uses.
We judge ads by outcomes, not vibes.
Some ads look “ugly” and sell. Some beautiful ads do nothing.
With Andromeda, the fastest way to lose time is to optimise for what feels right instead of what converts.
The future skill stack is less “I can build a complicated targeting setup” and more “I can produce and test messages people actually respond to, then measure it properly.” That is a more creative job, and honestly, a more useful one.
