You've probably noticed that something feels off with your Meta ads over the past year.
Creative that used to work stopped working. Budget started consolidating in weird ways. One ad with mid results would eat 80% of spend while others sat dormant. Your Facebook and Instagram ad performance just got, for a lack of a better word, weird.
Meta's got a fancy, buzzy name for all of this…. Andromeda. You've seen it slapped onto the start of hundreds of Guru LinkedIn posts, heard it during calls with your Meta rep, and flagged by your agency. It's Meta's redesigned ads retrieval system, the stage before the auction where Meta's algorithm decides which ads even get evaluated for a given impression.
When Meta announced Andromeda, they made it sound like a major platform shift. In some ways, it absolutely is. But here's the kicker… the rollout started well before the public fanfare, and we, as well as most agency owners I talk to, had already adjusted our creative strategies months before Meta talked about Andromeda publicly.
As someone who runs a performance creative agency for DTC CPG brands, Y'all, I've spent quite a lot of time helping clients optimize their Meta advertising and adapt to these shifts. We don't have all the answers, and I'm skeptical of anyone who claims they do. But we're seeing patterns, and we're adjusting based on what's working.
Here's what we think is happening and how we're approaching Meta ad creative differently for DTC and ecommerce brands.
Here's the working theory, based on what Meta has said and what we've observed:
Meta is now analyzing the creative itself, imagery, tone, message, format, pacing, emotions, visual cues, and using that to match users with ads.
The system is also scanning at a much larger scale than before. Instead of evaluating thousands of creatives, it can process millions in real time. This is almost certainly being used to train Meta’s AI, but that’s a whole separate conversation.
What we do know is this: brands with small, repetitive creative libraries have been struggling for the better part of a year. Brands with larger, more diverse libraries have been seeing better results. Whether that's because of Andromeda specifically or just because Meta has been pushing for higher creative volume for a while now, hard to say. Probably both.
One of our clients in the health and wellness industry began to notice a decline in the discernible data coming out of headline tests. Everything was muddled. After we restructured testing to utilize completely visually unique creative variations, things improved and we got actionable data and clear messaging direction that we were able to pass to the client. Was that Andromeda? Was it creative fatigue? Was it seasonal? We can't isolate the variable perfectly, but the change worked.
So, instead of treating this like a crisis or a massive platform overhaul, we're treating it like an evolution. The behaviors we're seeing now have been building for months. In fact, we, like many agencies, have been advocating for creative diversity on accounts for a long while now. We continue to adjust our creative approach to match what the platform is rewarding.
Whether you call it Andromeda or just "how Meta works now," we've changed how we think about creative over the last year. Not because of a single announcement or particularly panicky post by your favorite marketing bro on X, but because we kept seeing the same patterns across client accounts.
Here's what we've moved away from and what seems to be working better.
We've reduced micro-testing.
Swapping a headline on a static ad, using the same structure but with a different background image, shooting the same script from a slightly different angle; we're doing less of all of these things. Not because ad testing wholesale doesn’t work anymore, but because these types of changes don't seem to register as meaningfully different to the Meta algorithm post-Andromeda.
When we run micro-variants, our suspicion is that they compete for the same impression pool rather than expanding reach. One, likely based on misleading initial data that Meta reads too much into, gets most of the spend while the others barely deliver. It's not that the underlying test was wrong, it's that the platform can’t seem to treat the candidates within the test as distinct options.
This was a hard shift for us initially. We pride ourselves on our rigorous A/B testing and our ability to bring our clients data and ad revenue when they work with us. But if the system is clustering similar creatives together, then isolated A/B comparisons don't tell us much anymore.
We've stopped producing high volumes of sameness.
Five tweaked versions of the same static ad structure don't seem to equal creative variety in Meta's eyes. Neither do ten mash-ups that all use the same visual rhythm and opening style. Ad creative diversity matters more than raw volume.
There are some tools coming out that approximate what Meta sees as unique. We've recently been playing with Recharm's Asset Similarity tool for this. Is it perfectly accurate? Probably not! But it does give us a sense of how similar an AI algorithm may perceive two ads to be.
I want to emphasize that we are not advocating for the complete exclusion of headline testing and use of micro-variants. The need for creative diversity even in testing is, however, a consideration we take into account when determining ad mixes and when assessing performance. Welcome to the gray area that will be advertising on Meta ads in 2026.
We've started building more concept-level variety.
Now, when we build Meta ad creative, we focus on making sure each ad differs across core angle, persona, format, or visual identity, while keeping enough consistency to compare results and pull statistically significant data. As I mentioned earlier, this is tough and has taken us some time and lots of testing to refine. We now categorize creative as Andromeda-optimized or not.
In addition to variant-level changes, we now also push for more distinction between our concepts. What in the past would have been three problem-solution concepts that each performed well by most standards is now problem-solution vs. transformation vs. founder POV vs. attention-grabbing listicle. This approach to Facebook ad creative optimization means each unique concept gets multiple executions, but the executions stay visually and structurally distinct.
Here is an example of what Andromeda-optimized variants look like for us. The same base concept, with distinctly different visual presentations.
(IMAGE COMING)
This has had a huge impact on how our creative strategists create concepts and our designers and editors produce creative. In full transparency, it was kind of a mess at first. However, we've refined our approach and have dialed in what it means to create ads that are both unique and yet also fit within a system that allows for learnings to be extracted.
Fortunately, it's paid off so far. In our day-to-day experience, accounts with multiple distinct concepts perform better than accounts with repetitive creative, all else being equal.
I'm hesitant to say anything definitively here because there are so many variables at play. Seasonality, creative fatigue, market conditions, and budget changes make it hard to isolate what's driving results. But there are patterns we keep seeing across client accounts.
Bigger, more diverse creative libraries help.
For DTC and ecommerce brands, the key is that it's not just more ads, it's more kinds of ads. Different formats, personas, emotional territories, completely distinct layouts.
A creative testing system is still critical.
The brands performing well right now aren't just throwing spaghetti at the wall. They're tracking their creative by concept area, persona, and format. They’re relying on proper creative strategy more than ever before. They know which distinct concepts are working and which need refreshing. They have workflows for rotating in new executions and retiring underperformers.
We're measuring creative differently.
We're not ignoring ROAS and CPA, but we're also rarely making decisions solely based on per-ad ROAS. Unless forced, the platform doesn't distribute spend evenly, so comparing Ad A to Ad B in isolation doesn't always provide meaningful insights (depending on the account structure).
Instead, we're looking at directional metrics across concepts: CTR (click-through rate), hook rate, hold rate, and comment sentiment. We're seeing which concepts are consistently rising to the top, not just which individual ad won this week.
And we're moving faster. Creative seems to decay faster than it used to, whether that's Andromeda or just the nature of digital marketing in 2025, probably both. But we're adding new concepts weekly and constantly challenging ourselves to push the boundaries of what creative looks like for optimal Meta ad account performance.
If you're reading this and thinking your ad diversity isn't Andromeda-optimized, here's where I'd start.
Audit your creative library.
Pull everything that's currently active and group it by concept. Be honest about what actually looks and feels different versus what's just a micro-variant. Tools like DataAlly make this far less daunting than it used to be.
Use a tool like Recharm's to test ads against each other. If you've got 30 ads but only 5 real distinct concepts, that might explain why budget is consolidating so heavily and you're having trouble scaling your Meta ads profitably.
Identify which distinct concepts you're missing.
Which distinct concepts are represented in your ad mix? Are all of your ads just different versions of Us vs. Them? Are you overindexed on Problem-Solution? Missing out on founder-led content? Which personas are underrepresented? Which formats are you avoiding?
You don't need to fill every gap immediately, but having a plan to add 2–3 new distinct concepts over the next couple months will only benefit your account.
Consider expanding your creator bench.
If you're working with one or two creators, think about recruiting a few more. If you're not working with UGC creators at all, it's time to test it. This doesn't have to be expensive. We've been seeing amazing results from new creators (largely because they come off as real and authentic!). What matters is variety in tone, energy, and aesthetic.
Build a system for tracking and rotating creative.
You don't need anything complicated. Just make sure you can see your full library at a glance, understand what's working, and know when it's time to retire something or double down.
Give it time, but stay flexible.
If you make changes to your creative approach, give it a few weeks before you draw conclusions. The data takes time to become statistically significant and the impact of things like seasonality needs to be taken into consideration. If something clearly isn't working, turn it off. The platform is still evolving, and what works today might need adjustment in three months.
Truthfully, we might all look back in six months and realize we were overcomplicating this (very possible!). Or we'll have new information that changes how we think about creative strategy entirely. That's okay!
Here's what I'm confident about: Creative diversity and quality matters more than ever for Meta advertising success. Brands that treat creative as a strategic system to deploy diverse ads for a multitude of personas and across many angles are seeing better results and getting better data.
I don't think anyone outside of Meta truly knows how Andromeda work, and I'm skeptical of anyone claiming they've cracked the code. If you’ve talked to a Meta rep recently, you’ll see that they don’t necessarily understand what is going on either. What we have are observations, hypotheses, and a willingness to adapt when something stops working.
For us, that's meant investing more in creative diversity, reducing our reliance on micro-testing and evolving to new testing methodologies, building systems to track and rotate our libraries, and moving faster with creative production. It's what's working for our clients.
The DTC world has always rewarded adaptability. iOS updates, platform algorithm changes, shifting consumer behavior, tariffs… we've navigated all of it. This feels similar. Not a crisis, just another evolution that requires us to pay attention and adjust. For those willing to embrace creative diversity, this presents a huge opportunity to optimize to Meta’s algorithm while many remain stuck in old patterns.
If your creative strategy hasn't changed in the last year and you're still seeing good performance, keep doing what you're doing. Start to fold in new types of creative, but don’t turn off a good thing. If things have been harder lately and you're not sure why, hopefully some of the above is useful.
If you want to talk through your specific creative strategy or compare notes on what you're seeing, email me at travis@yall.co.
What is Meta Andromeda? Andromeda is Meta's redesigned ads retrieval system that determines which ads get evaluated for each impression before the auction even happens. It analyzes creative elements like imagery, tone, format, and pacing to match users with relevant ads.
When did Meta Andromeda roll out? While Meta officially announced Andromeda recently, the system has been gradually rolling out for 6-12 months. Many advertisers noticed changes in their Facebook and Instagram ad performance throughout 2024 before the official announcement.
Does Andromeda mean audience targeting is dead? No. Audience targeting still exists, but it functions more as guardrails now. Creative itself has become the primary signal Meta uses to match ads with users.
How many creatives should I have in my Meta ad library? Focus on concept diversity rather than raw numbers. Most successful DTC brands maintain many distinct concepts with multiple unique executions.
What's the most important part of an ad under Andromeda? The first three seconds. Andromeda makes retrieval decisions quickly, so ads need immediate context, visual contrast, and emotional clarity right from the start.
Discover new D2C brands, new eCommerce tools and read in-depth founder reviews each week.
Everything you need to build a better D2C brand, business and tech stack.
