Quick Decision Framework
- Who This Is For: Shopify founders and marketers doing $10K to $2M per month who are running Meta ads and have noticed CPMs climbing, ROAS dropping, or campaigns that worked six months ago suddenly underperforming with no obvious explanation.
- Skip If: You are not running paid ads on Facebook or Instagram yet. Come back when you have an active Meta ad account and at least 90 days of campaign history to compare against.
- Key Benefit: Understand exactly why Meta’s delivery system changed, what the new rules are, and how to restructure your creative strategy so your ad spend starts working again within 30 to 60 days.
- What You’ll Need: Access to your Meta Ads Manager, at least one active campaign, and a willingness to rethink how you produce and test creative. A basic understanding of Advantage+ campaigns is helpful but not required.
- Time to Complete: 9 minutes to read. 2 to 4 hours to audit your current account structure and begin building your first new creative pack using the framework in this article.
Your CPMs did not go up because the ad market got more competitive. They went up because Meta’s system is telling you something specific: your creative does not match any audience it can find. That is a very different problem, and it has a very different solution.
What You’ll Learn
- What Meta’s Andromeda system actually is, why it was built, and why the timing of its rollout explains the performance drop so many Shopify brands experienced starting in mid to late 2025.
- Why the audience targeting playbook that drove results for the past five years is now actively working against you, and what replaced it.
- How Andromeda uses your creative as the targeting signal, and what that means for how you brief, produce, and test ads going forward.
- Why running five versions of the same ad with different headlines is not creative diversity under the new system, and what actually counts as a distinct signal.
- The specific account structure and creative approach that is working for Shopify brands right now, broken down by stage so you can apply it at your current scale.
Why Your Meta Ads Stopped Working (And It Is Not Your Offer)
Starting in the second half of 2025, a pattern showed up across Shopify ad accounts that had nothing obvious in common. Brands with proven offers, strong creative, and years of profitable Meta history started watching their CPMs climb and their ROAS compress. The accounts that got hit hardest were often the ones that had been running the most sophisticated targeting setups: layered interest stacks, lookalike audiences built from purchase data, retargeting sequences with custom exclusions. The more complex the structure, the worse the performance got.
The explanation that most agencies gave their clients was some version of “the ad market is more competitive right now.” That is partially true and almost entirely unhelpful. The real explanation is that Meta rebuilt the core engine that decides which ads get shown to which people, and the old playbook is not just less effective under the new system. It is actively penalized by it. The update is called Andromeda, and understanding what it actually does is the difference between spending the next six months chasing your old ROAS and building a system that works with the new rules instead of against them.
Andromeda is Meta’s AI-driven ad retrieval engine. Before any auction happens, the system has to narrow down tens of millions of active ads to the few thousand it will actually consider showing a given user. That filtering step is retrieval, and Andromeda is what runs it. Meta’s engineering team described it as enabling a 10,000x increase in the complexity of the models used for ads retrieval. In plain language: the system is exponentially better at matching creative to users than the old system was, which means it no longer needs your manual audience settings to do that job. And when you give it manual audience restrictions anyway, you are constraining a system that is smarter than your targeting setup.
What Andromeda Actually Changed About How Ads Get Delivered
The old Meta delivery model worked roughly like this: you defined an audience, Meta found people in that audience, and your ad competed in an auction for their attention. Your targeting was the primary signal for who saw your ad. Creative mattered, but it was secondary to audience selection. The brands that won were the ones that built the most precise audiences and tested creative variations against those audiences systematically.
Andromeda inverts that relationship. Under the new system, your creative is the primary targeting signal. The ad itself tells the algorithm who it is for. A founder-led video talking about the problem a product solves signals “show this to people who have expressed that problem.” A UGC testimonial from a 35-year-old woman signals “show this to people who resemble her and respond to social proof.” A product demo focused on a specific use case signals “show this to people in that use case.” Meta’s system reads those signals and matches the ad to the right user without you having to define the audience manually. The more distinct and specific your creative signals are, the more precisely the system can match them.
This is why the brands that got hit hardest were often the ones with the most complex targeting setups. They were giving Andromeda conflicting instructions: their manual audience settings said “show this to women 25 to 44 who like yoga and have purchased athletic wear in the last 30 days,” but their creative was a generic product shot that could apply to anyone. Andromeda looked at the creative, could not find a strong signal, and either skipped the ad in retrieval or charged a high CPM to force delivery into an audience the system was not confident about. The CPM spike is not a market condition. It is the system’s feedback signal telling you the creative does not match anyone it can find with confidence.
Why Creative Diversity Is Now The Core Skill
The most common misreading of what Andromeda requires is that brands need more creative volume. That is not quite right, and chasing volume without understanding what counts as a distinct signal will waste your production budget faster than the old system ever did. Andromeda’s retrieval engine groups similar ads together. If you run five versions of the same ad with different opening lines, the system sees them as the same signal repeated five times. It does not treat them as five separate creative tests. It treats them as one weak signal with redundant copies, and it spreads your budget thin across all of them without learning anything useful.
What Andromeda rewards is creative that speaks to genuinely different buyer states, motivations, or moments. A problem-aware buyer who has been struggling with a specific issue for months responds to a different creative than a solution-aware buyer who is comparing options. A first-time buyer in the consideration stage responds differently than a lapsed customer who bought once and did not return. Those are different people with different contexts, and the creative that reaches them needs to signal those different contexts clearly. Changing a headline or swapping a product image does not do that. Changing the angle, the hook, the format, and the emotional register does.
The practical implication is that your creative brief process needs to start with buyer state, not product feature. Before you brief a new ad, ask: what is this person feeling right now, and what does the creative need to say to match that feeling? A brand selling a sleep supplement might need a creative for the person who has tried everything and is exhausted and frustrated, a different creative for the person who is newly interested in sleep optimization and wants to understand the science, and a third creative for the person who bought once and did not build a habit. Those are three distinct signals. Three versions of a product shot with different headlines are not. For a broader look at how AI systems are reshaping how ecommerce brands reach buyers across every channel, the AI agents in ecommerce overview connects the dots between what is happening inside Meta and what is happening across the broader discovery landscape.
The Account Structure That Works Now
The structural implication of Andromeda is almost the opposite of what most Shopify brands have been building for the past five years. The high-performing account structure right now is simpler, not more complex. Fewer campaigns. Fewer ad sets. More creative variation at the ad level inside a small number of broad ad sets. The system does the audience work. Your job is to give it enough distinct creative signals to work with.
The setup that is working for most ecommerce brands right now looks like this: one campaign per objective, one to two broad ad sets with minimal audience restrictions, and a rotating pack of six to twelve genuinely distinct creatives inside each ad set. Advantage+ Shopping Campaigns are purpose-built for this structure and are worth testing if you have not already. They hand audience selection and placement optimization entirely to the algorithm and let you focus entirely on creative input, which is exactly what Andromeda rewards.
Budget consolidation matters too. Andromeda needs enough data to complete its learning phase and optimize across the signals you are giving it. Micro-budgets spread across many campaigns starve the system. A consolidated budget in two to three campaigns gives the algorithm enough volume to learn quickly and allocate spend to the creative signals that are working. The threshold varies by account, but the general principle is that fewer, better-funded campaigns outperform many thin campaigns under the new delivery model. If your current account has more than five active campaigns targeting similar objectives, that is the first thing worth auditing.
What Clean Data Has To Do With All Of This
Andromeda’s matching precision depends entirely on the quality of the conversion signals you are sending back to Meta. The system learns which creative signals lead to purchases by tracking what happens after someone clicks. If your pixel data is incomplete because of iOS privacy restrictions, ad blockers, or a misconfigured setup, the system is learning from a distorted picture. It thinks certain creative signals lead to purchases when they actually lead to abandoned carts, or it misses purchase events entirely and cannot credit the creative that drove them.
The fix is Conversions API, or CAPI, running alongside your pixel. CAPI sends conversion data directly from your server to Meta, bypassing the browser restrictions that block pixel data. Running both in parallel gives Meta the most complete possible signal about what is actually converting. This is not a nice-to-have under Andromeda. It is foundational. If your CAPI is not set up or is sending incomplete event data, you are asking the system to optimize creative matching based on partial information, and your results will reflect that gap regardless of how good your creative is. Shopify’s native Meta integration includes a CAPI setup path, and it is worth verifying that purchase events are firing correctly with full event match quality scores before you make any other changes to your account.
The Creative Refresh Reality Most Brands Are Not Ready For
One of the harder truths about Andromeda is what it does to creative fatigue timelines. Under the old system, a strong creative could run for months before performance degraded meaningfully. Under the new system, ad fatigue sets in after roughly seven to eight exposures per user, and because Andromeda’s broad targeting reaches users more efficiently, your best creative burns through its audience faster than it used to. A creative that would have run for three months in 2023 might be fatiguing in four to six weeks in 2026.
The brands that are managing this well have built production systems, not just creative assets. They are briefing new creative packs on a weekly or biweekly cadence, retiring creatives when CTR drops below 1.5% or CPM jumps 25% week over week for three consecutive days, and treating creative production as an ongoing operational function rather than a periodic campaign project. For most Shopify brands doing under $500K per month, that does not mean producing 50 creatives a week. It means having a clear brief template, a reliable production workflow, and a decision rule for when to retire and replace. The brands getting crushed are the ones still treating creative production as something that happens quarterly when the campaign “needs a refresh.”
The connection between this shift and the broader changes happening in how buyers discover and evaluate products is not coincidental. Meta is moving toward the same AI-driven matching logic that is reshaping organic discovery through agentic commerce. In both cases, the signal that gets you found is the specificity and clarity of what you put into the system, not the precision of your manual targeting or bidding strategy. For context on how that shift is playing out across the full discovery landscape for Shopify merchants, the agentic commerce guide for Shopify 2026 covers the structural picture in detail. And if you want to understand how AI shopping agents are starting to influence the buyer journey before they ever reach a Meta ad, the rise of AI shopping agents piece connects those dots directly.
Where To Start If Your Account Is Underperforming Right Now
The fastest path to recovery is not a new campaign. It is an honest audit of what you are currently giving the system to work with. Start by pulling your active creatives and asking a single question for each one: what buyer state does this creative signal? If the answer is “anyone who might want this product,” the creative is not giving Andromeda a usable signal. It will either not make it through retrieval or it will deliver at a high CPM because the system cannot find a confident match.
From there, identify two or three genuinely distinct buyer states for your top product. Brief one creative concept for each state. Not variations of the same concept. Different angles, different formats, different emotional registers. A problem-aware hook, a social-proof testimonial, a founder-led explanation. Put those into a single broad ad set inside an Advantage+ Shopping Campaign with your full purchase budget consolidated behind it. Let it run for at least two weeks without structural changes. Watch which creative signals are getting budget and which are getting skipped. The ones getting budget are the signals Andromeda can match. Double down on those angles and retire the ones that are not getting spend.
That is the feedback loop the new system is designed to give you. The brands winning on Meta right now are not the ones with the biggest budgets or the most sophisticated targeting setups. They are the ones who figured out that Andromeda is a signal-matching engine and started treating creative production as the primary lever for growth. For a deeper look at how AI is transforming ecommerce operations beyond paid acquisition, including how the same signal-matching logic is reshaping retention, personalization, and discovery, the agentic commerce operations piece is worth reading alongside this one.
Frequently Asked Questions
What Is Meta Andromeda And Why Did It Affect My Ad Performance?
Meta Andromeda is the AI-driven retrieval engine that filters tens of millions of active ads down to the few thousand Meta considers showing any given user before the auction even happens. It was built to handle the explosion in ad volume driven by AI-generated creative and to enable more precise creative-to-user matching at scale. The performance impact for Shopify brands comes from a fundamental shift in how the system works: under Andromeda, your creative is the primary targeting signal, not your audience settings. Brands running complex manual targeting setups found their CPMs spiking because they were constraining a system that no longer needed those constraints, and their generic creative was not giving the algorithm a strong enough signal to match confidently.
Why Are My CPMs Going Up Even Though I Have Not Changed Anything?
Rising CPMs under Andromeda are almost always a creative signal problem, not a market problem. When Andromeda cannot find a strong match between your creative and a user who is likely to convert, it either skips your ad in retrieval or charges a high CPM to force delivery into an audience it is not confident about. The CPM spike is the system’s feedback signal telling you the creative does not match anyone it can find with confidence. The fix is not to increase your budget or tighten your targeting. It is to give the system more distinct, specific creative signals that it can match to identifiable buyer states. If your CTR is also dropping alongside rising CPMs, that confirms the creative is reaching the wrong people, which is a retrieval and matching issue, not a bidding issue.
What Does Creative Diversity Actually Mean Under The New Meta System?
Creative diversity under Andromeda means ads that signal genuinely different buyer states, not ads that look slightly different from each other. Running five versions of the same product shot with different headlines is not creative diversity because Andromeda’s retrieval engine groups similar ads together and treats them as the same signal repeated. Real diversity means different angles, different formats, different emotional registers, and different hooks that speak to different moments in the buyer journey. A problem-aware hook, a social-proof testimonial, a comparison-focused demo, and a founder-led explanation are four distinct signals. Four product images with different background colors are not. The practical test is whether each creative could plausibly be described as speaking to a different person in a different state of mind.
Should I Still Use Audience Targeting Or Switch Entirely To Broad Targeting?
For most Shopify brands, broad targeting with Advantage+ Shopping Campaigns is the right default under Andromeda. The system is designed to handle audience discovery through creative matching, and manual audience restrictions constrain that process without adding meaningful precision. The exception is retargeting, where you have a specific list of people who have already interacted with your brand and you want to reach them with a specific message. Even there, the trend is toward letting Andromeda handle the matching within a defined pool rather than building complex exclusion and inclusion stacks. The general principle is to give the system as much room as possible to find the right match for each creative signal, and to focus your energy on making those signals as distinct and specific as possible.
How Often Should I Be Refreshing My Creative Under Andromeda?
The refresh cadence depends on your budget and reach, but the general benchmark is to monitor for fatigue signals every week and refresh when CTR drops below 1.5%, CPM jumps 25% or more week over week for three or more consecutive days, or ROAS falls for three or more consecutive days without an obvious external cause. At most Shopify scales, that translates to a biweekly creative review and a new creative pack every two to four weeks. The brands managing this well have built production systems rather than treating creative refresh as a reactive task. That means a brief template, a production workflow, and a clear retirement rule so that creative decisions are made on data rather than gut feel about when something has “run long enough.”
