ArticleIssue #144 | Why Your Meta Ads Are Struggling (Even With Great Creative)
I hope you all had a relaxing, restful, turkey-filled Thanksgiving. For those of you living in the eComm trenches, I hope your BFCM is humming…or at least not entirely melting down.
If you’ve spent even five minutes on Marketing Twitter over the last 48 hours, you’ve probably seen the usual chaos: some brands posting record-breaking screenshots, others quietly panicking over higher-than-expected CAC and erratic performance. The vibes are always mixed.
There’s no doubt this week is the eComm Super Bowl – and just like in football, the outcome doesn’t usually come down to trick plays or hero moments. It comes down to fundamentals.
Last week, I audited a Meta ad account for a fast-growing sporting goods brand. This was an unusual audit, primarily because this brand was nothing like the usual ones I review. They had partnerships most brands couldn’t imagine in their most fever-ridden dreams, a remarkable founder story, exceptional creative, strong product-market fit and a loyal customer base…yet it was posting net operating losses in excess of 30%.
Think about that for a minute: this brand was doing everything most marketers say to do: create exceptional, world-class products that people actually want? Check. Forge great partnerships with established, category-adjacent brands? Check. Build a stable of influencers who love and support the brand? Check. Create emotive, resonant content that shares a credible story? Checked so hard Shopify featured them in national commercials.
And somehow, the brand was not just losing money, they were hemorrhaging it.
Once I got into the account, I found the reason wasn’t mysterious; it’s something I see in roughly 80% of the audits we do: the problem was the structure.
You can have the best product, killer creative and exceptional storytelling, but if your account structure is actively working against Meta’s algorithm, you’re (at worst) unintentionally kneecapping your growth or (at worst) actively setting money on fire.
Anyone who has run Meta Ads for the last ~5 years has heard the non-stop refrain that “Creative is Key” and “Creative is the targeting” – but (in true Hegelian fashion), the creative fever dream is breaking and more and more brands are starting to realize that great creative can not out-run broken fundamentals.
The difference between a record year and a midseason meltdown is almost always structural, not superficial.
Four Silent Killers That Derail Meta Performance
1. The Attribution Trap
The first place I look in a Meta audit: attribution settings.
In this case, the account was heavily leaning on 1-Day View. Nearly half the reported conversions were people who never clicked an ad – Meta just claims those people saw it, then bought later via some other channel.
In those cases, Meta is neither creating the demand nor is it capturing demand. It’s just there, collecting a tax, then using that tax collection as justification for its value.
What’s worse? If you’re scaling spend based on that view-based conversion data, you’re not scaling customer acquisition; you’re inflating performance based on people who had an exceptionally high probability of purchasing anyway.
Fix it: Shift to 7-Day Click (7DC) attribution. Yes, the numbers will look worse short-term. But they’ll be real. Meta’s algorithm will start optimizing for people who stop, click and purchase.
I’ve found that 1DC tends to over-index toward high intent users (especially for mid-to-high consideration purchases), whereas 1DV tends to take credit for far too many non-incremental conversions. 7DC is a nice balance between the two, where Meta is still incentivized to bring new-to-brand buyers (avoids the 1DC trap) while still forcing it to add real value (the 1DV trap).
2. The “Peanut Butter” Segmentation Mistake
I’m constantly amazed at how few brands understand the damage advertising products or services with wildly different unit economics into the same campaign can do.
In this account, high-margin accessories were running alongside high-CPA hard goods and mid-margin apparel.
To put a finer point on this: the accessories had margins of 76%+ – for every $100 sale, the brand pulled in about $76.70 after costs of goods sold + cost of delivery (including payment processing fees, shipping, allowances for returns, customer service, packaging – all of it). The hard goods (think: shoes) had a radically different margin profile – less than 24% after partnership commissions, COGS, R&D allowance, shipping, returns, customer service, packaging, etc. And in the middle was apparel – margins were 40% to 60%, depending on the specific piece – but return rates were MUCH higher than any other category.
That put the brand in an impossible position when it came to assessing performance. The old agency raved about the 3.00 ROAS – but neglected to mention that 75% of spend went to the hard goods (which would require a 4.00 ROAS just to break even). The rest was split evenly between accessories + apparel. End result? The account was running at a pretty substantial loss, even though the ROAS numbers and cost/acquisition looked solid – all because the structure wasn’t set up to facilitate optimization that helped the business.
The above is reason enough to properly segment your campaigns; but that isn’t even the real problem. The true problem is that this setup simultaneously under-spends and over-spends – which means the brand gets the worst of both worlds:
- Accessories, which could be run profitably at a 1.5x ROAS, misses out on thousands of customers because the ROAS targets are too high / CAC targets are too low
- Hard Goods, which should be run at a 5.00x ROAS (sport-specific shoes aren’t exactly an LTV play), are allowed to run at a 3.50x – which means every sale costs the brand ~$65
- Apparel gets stuck in no-man’s-land, where it doesn’t have the budget to scale and doesn’t have the margins to thrive at low volumes
Put all of that together, and this brand is simultaneously missing out on a massive segment of their addressable audience, all while losing money on every sale.
You might be wondering how the first part is possible – well, here’s the audience/target graph. It exists for every brand, every SKU, every product:
The underlying principle is this: as you make a target less restrictive (lower ROAS, higher CPA), the available audience on the platform increases exponentially. That’s not some marketing bullshit; that’s (semi-advanced) math.
The second core consideration: ad platforms are like water – they will always choose the path of least resistance. If you lump cheap socks and heritage polos together with a single target, Meta (or Google, or Snap, or TikTok) is going to sell socks, simply because they’re cheap and convert faster. That starves your core, high-margin products.
Fix it: Segment campaigns by margin profile.
- Campaign A: Accessories (high margin, lower CPA)
- Campaign B: Hard Goods (lower margin, higher CPA)
- Campaign C: Apparel (mid-margin, mid CPA)
Then set tROAS or cost cap targets that reflect each category’s profitability. This lets the algorithm optimize based on actual outcomes, not a blended efficiency target that hides what’s really going on under the hood.
3. The “Lazy Broad” Targeting Fallacy
We’ve all heard for years that “broad is best” or “the creative is the targeting” – and this brand (along with many others) took that message to heart.
Now, before I hear from all the Meta Ads people that broad is great: there’s a version of broad targeting that is both powerful and appropriate. That’s the one where the algorithm is given wide surface area while it is being fed clear, high-quality signals AND the advertiser has a staggeringly large addressable market.
When all of those conditions are not true, broad targeting is just signal dilution at scale, masquerading as a “strategy.” It’s the kind of undifferentiated reach strategy that looks “simple” on paper but collapses the minute you inspect the economics of the business.
For this brand, 80% of revenue is generated by men aged 35–54. Why? Because roughly 78% of the people who play the sport this brand is built around are men, aged 28-62 – mostly from affluent or semi-affluent households.
It isn’t an exaggeration to say that this audience is the gravitational center of the P&L.
These are the customers who understand the product, respond to the value prop and convert at sustainable rates. Broad targeting would make sense after Meta has been told that this cohort is the North Star – but instead, the account was treating them like one of dozens of interchangeable options.
Meta was being asked to optimize a performance engine without being given the most important instruction: “These are the people who actually buy.”
To make matters worse, the vast majority of the remaining 20% of sales coming from women were not primary purchasers. They are gifters: spouses, partners, family members who were buying not because they resonated with the product, but because the product resonated with the men in their lives.
And yet, the account lacked even the most basic gifting creative: no “Perfect for him.” No “Voted #1 New Brand by [Insert Publication Here].” No messaging that gives a non-user confidence that they’re choosing a brand/product that will bring a smile to the face of their loved one.
Without that contextual scaffolding, the female audience did what misaligned audiences always do: drastically underperformed relative to the core target.
This created a predictable, dangerous dynamic:
- Meta naturally chased cheap impressions
- Cheap impressions tended to skew female or younger
- The gifting audience wasn’t given a compelling angle
- The male buyer wasn’t given enough volume or message precision.
End result? Spend moderately evenly distributed across gender + age demographics. CAC rising as spend to non-performing groups rose disproportionately to sales. Younger “aspirational” audiences purchase lower-value items (polos, shirts, socks, cheap accessories), which further confuses Meta.
In other words, the problem wasn’t that broad targeting “didn’t work” – it was never going to work! When 80% of your revenue is concentrated in one segment, broad targeting is not a strategy; this brand requires clearer prioritization from the account structure for Meta to be effective.
Aside: How To Ruin A Good Lookalike
That was far from the only problem: this brand also decided to run multiple, wildly-different lookalikes in the same ad set: a 10% LAL of all visitors along with a 3% LAL of “high value” purchasers. You can see how this is a problem when 75% of the brand’s traffic came from Meta, most of which came from the aforementioned broad targeting campaign:
- Broad targeting drove thousands of non-ICP visitors to the site, the overwhelming majority of whom did not convert
- The LAL comprised of site visitors was increasingly comprised of those who did not buy
- 76.9% of the combined (3% High Value LAL + 10% Visitors) audience was comprised people who were not the ICP – essentially turning a LAL into a Lookalike Broad which…isn’t a thing.
This undercut the entire premise of a Lookalike Audience is to reach net-new people who are behaviorally similar to an existing, high-value audience. That signal is the advantage – you’re using your business data to compress the surface area that Meta has to search to find buyers, with the goal of making acquisition more cost-effective because you’re limiting the search to the highest-probability areas.
Think of it like me tasking you with finding football cleats at a University. You could conduct a thorough, uniform search of the entire campus – every building, every classroom, every dorm, every dining hall – but that would be wildly resource-intensive. The smarter solution would be for me to tell you: the places where football cleats are MOST LIKELY to be found are in the football locker room, the training rooms, the football stadium and the football coaches offices.
Are there cleats in dorms? Yep. Are there cleats in the dining halls? Probably! But the probability of finding cleats in either location is infinitesimally smaller than the probability of finding them in the high-likelihood places, so direct your resources there. Fish where the proverbial fish are.
When you mix high-intent, high-probability signal with random noise, you degrade the quality of both signals. Targeting a 3% LAL of the brand’s best customers was a very good strategy. Mixing it with a 10% LAL of all website visitors – knowing that the majority of those visitors come from broad targeting on Meta – was a terrible strategy. The end result wasn’t mediocre, it was horrible.
Combine the two errors together and the inevitable result is underperformance across the board.
4. The Learning Phase Math Problem
Most teams talk about the Learning Phase as if it’s a mysterious black box – a volatile, unpredictable state where performance oscillates for reasons nobody fully understands. But in practice, the Learning Phase is not a philosophical problem or a platform quirk; it’s a basic optimization problem that many advertisers keep trying to solve emotionally instead of mathematically.
Meta’s delivery system is probabilistic. It needs a certain number of conversions per week – historically around 50 per ad set (now sometimes less) – in order to build the algorithmic confidence required to make stable, high-performing decisions. If you fund an ad set in a way that makes that threshold mathematically impossible, the algorithm cannot exit Learning. This is not because the platform is malfunctioning, but rather because the conditions for learning were never met.
This brand had built an architecture full of underfunded ad sets, each one operating at a budget level that made stable delivery impossible. For a group of ad sets, the target CPA was north of $150, but the budget was $50/day. How in the world do you get to 25 conversions (let alone 50) at $150/conversion (requires $3,750 per week in spend) when your budget is capped at less than 10% of that?
This configuration failed to provide Meta the financial runway to accumulate sufficient conversion data to learn anything useful. The algorithm was being asked to operate with insufficient data density, so it did what underfed machine-learning systems do: fluctuated wildly, overcorrected constantly and explored new audiences constantly.
The worst part? This was not an optimization issue or a Meta bug. The system was doing (mathematically) the right thing.
Underfunded ad sets don’t just underperform – they destabilize the entire account. It’s a vicious cycle that goes like this: budget fragmentation amplifies variance. Variance triggers algorithmic overcorrection. Overcorrection increases CPMs. Higher CPMs reduce impression volume + net-new reach. Lower impression volume + lower reach combine to weaken signal density further.
The account gets caught in a death loop, where the system can’t do what it’s being asked to do, so (like a poker player on tilt) it just takes riskier and riskier bets, hoping to strike gold that doesn’t exist.
The fix is as unglamorous as it is non-negotiable: fewer ad sets, funded properly.
When you consolidate spend, you give Meta a sufficient volume of conversion data to generate patterns. When you move to CBO (with clear objectives), you let the system distribute budget toward the pockets of efficiency it discovers. When you align the budget to SKU-level CPA/ROAS, you give the algorithm the environment required to exit Learning, which is the entire point of machine learning.
Your structure should be geared to supporting Meta in exiting Learning as quickly as possible, which means properly funding each ad set/campaign. Ultimately, this isn’t a philosophical debate, it’s a math problem. Until the math makes sense to the machine, nothing else will.
5. CPMr + Declining Marginal Reach: The Silent CPA Multiplier
If there’s one pattern that showed up across almost every audit in the last 6 months, it’s this: the accounts that struggled the most weren’t failing because of bad creative or targeting mistakes. They were failing because their marginal cost of reach had quietly exploded.
The metric responsible – CPMr (cost per 1,000 unique accounts reached) – is almost never tracked, and it certainly wasn’t even on this brand’s radar. But CPMr is the metric that tells you whether you’re actually reaching new people or simply re-serving the same impressions to the same audience segments over and over again.
That matters for one very important reason: when CPMr rises, CAC soon follows.
This brand is a perfect case study:
- Their 7-day frequency was above 2.8
- Their CPM had climbed above $105 (a level you typically only see in regulated categories like weight loss or gaming/sports betting)
- Meanwhile, competitors across their product lines were sitting with CPMs between $25-$50.
This brand had not done anything crazy – no material changes to targets/bids. No budget increases. No change to creative rotation/new creative cadence.
They weren’t “being penalized” by Meta; they were just caught in a bimodal audience distribution: Meta was serving some segments this brand’s ads 20x per 7 days (essentially hammering the remarketing/website/past buyer audiences), while frantically throwing darts in the dark in prospecting. The result was a sky-high average frequency, along with declining marginal reach among their ICP (detailed above) and frustration from their existing customers / past site visitors.
Because this brand was not proactively managing reach, Meta defaulted into a high-cost equilibrium where the system continued to serve the same audiences with the same ads over and over again, essentially buying the same attention 10x, while averaging down with low-quality, low-probability impressions.
This brand’s marketing team misdiagnosed this three times: first, as a creative problem (“ads are fatiguing”); second, as a targeting problem (“The audience is too narrow – we should go broader”); and finally as a budget problem (“We need to scale down budgets and reset”). What’s amazing is that they were wrong 3x in the same deck – that does take some skill.
But the root cause in this situation is almost always the same: declining marginal reach.
Reach declines → New Visitor % drops → CPMr rises → CAC increases.
The solution isn’t to panic or retreat. The solution is to rebuild the top of the funnel intentionally, not by accident and not through last-minute tinkering.
High-performing brands use a structured TOF strategy that includes:
- A dedicated Awareness campaign (always present)
- A secondary Reach or Video View campaign (for coverage)
- Strategic upper-funnel targeting (custom placements, interests, LAL splits)
- Creative variety across formats, tones and awareness levels
- Measurement frameworks that include incremental signals, not just in-platform metrics
This reintroduces cheap, high-quality reach into the system. When you reduce CPMr, you expand the top of your acquisition funnel, lower CAC and increase the overall surface area for conversions.
Upper-funnel investment is not fluffy or optional. When built correctly, it is the economic engine that stabilizes everything below it. That’s doubly true for brands (like this one) seeking to establish and maintain premium price points – most people aren’t going to buy a $300 pair of shoes the first time they see them. A $500 sporting goods accessory isn’t an impulse purchase for most Americans, especially in current economic conditions.
The brands that scale profitably year after year are rarely the ones making waves on X with clever hooks or massive new creative volume. Rather, it’s the brands that treat reach as a strategic asset, and CPMr as a leading indicator of future performance. Introduce your brand to relevant new potential buyers. De-risk the purchase for them via social proof, testimonials, repetition, compelling angles, relevant use cases. Capture them efficiently at or below target.
It’s really that simple – but it all starts with understanding that today’s buyers are rarely the ones that heard about your product for the first time an hour ago. Invest more in reaching the right new people, and the sales/efficiency will follow.
Meta’s Quiet Transformation: Creative Still Matters, But Not Like It Used To
Here’s what I wrote in Issue 137: Creative in the Age of Andromeda:
We’re not just in the “Creative is the Variable” era. We’re in the age where Meta’s system is beginning to determine creative value upstream, not just based on output (clicks, conversions) but on creative diversity, structural intent and audience interaction.
Meta’s algorithm is now proactively analyzing:
- How often you recycle the same assets
- Whether ads are “too similar”
- Creative cluster performance (fatigue, conversion rates, half-life)
In other words, the platform isn’t waiting to see which ad performs. It’s trying to predict whether your creative mix will work at scale, before it even delivers.
That means volume for volume’s sake no longer works. You can’t just throw 40 variants into the account and assume something will pop. Meta wants structured, intentional inputs. Part of that is because Meta genuinely does want to provide a superior experience to its users; most of that is because it wants more diverse data to train its own ML algorithms so it can charge higher prices for ad space going forward. If you don’t trust Zuck’s good intentions, trust his capitalistic imperatives.
Put another way: Meta wants creative diversity that trains the algorithm, not just fills space.
So what does that mean for how we approach creative?
1. Structure Dictates Creative Context
If your campaign architecture, attribution, and budgets are off, your best creative will underperform. Fix the machine first. Then focus on content. If – like this brand – you have a sizeable gifting audience: build that audience first, then ensure the creative in that ad set is targeted to that audience (i.e. different hooks, different messages, more de-risking language, more social proof).
2. You need creative variation that maps to different demand pockets.
Not just A/B testing headlines, but real diversity: different styles, tones, formats, hooks, angles. The algorithm thrives when it sees clear creative signals it can match to micro-segments inside broader audiences.
3. Structure is how you communicate priorities to Meta
Meta’s AI is filtering content based on fatigue, overlap, and relevance. If your assets are too similar, your delivery will be throttled before it begins. Creative is no longer just the message; it’s part of the infrastructure.
Second: structure is how you communicate what’s important to Meta. Ad Sets with decent budgets but sky-high ROAS targets? Nice-to-have. Ad sets with lower tROAS + sky-high budgets? Scale city. No new creative + Low CPA + Low Budgets? Ignore. You get the idea.
Most of this issue has focused on DTC/eCommerce – but the reality is that every principle here applies equally to B2B + B2C lead generation. Most B2C lead gen businesses have different service lines / product lines with wildly different margin profiles. Every one has core + non-core audience segments. All benefit from the same fundamentals discussed here.
Where Tools Like Optmyzr Make This Easier (and Smarter)
Juggling multiple Meta accounts for clients or business units quickly becomes overwhelming.
This is where Optmyzr for Social becomes a legitimate force multiplier.
It gives you:
- Cross-account budget controls to monitor pacing and efficiency at scale
- Rule-based automation to pause underperformers (e.g. high CPA, low CTR) before they waste spend
- Audience and creative structure tools to enforce segmentation and creative diversity
- Custom alerts and clean reporting so you’re not chasing performance manually
I’ll be the first person to say that fancy tools can’t (and shouldn’t) replace strategy – but when your strategy is sound, tools like Optmyzr give you the structure and leverage to execute it at scale, without burning your team out in the process.
Check out what Optmyzr can do for Meta ads and get an extra 15% off your annual subscription (30% + 15%) when you sign up by Dec 3.
The brand I audited has the potential to scale to $500k+//month, profitably. All of the pieces are there – product, story, creative, audience, influencers. What’s holding them back are structural issues and philosophical misunderstandings.
Creative is the biggest, most powerful lever for scale…but scale only happens when the system underneath is strong enough to support it.
If your Meta performance feels erratic this week, don’t start by tweaking copy or loading more hooks. Start by zooming out:
- Are you optimizing for actual action or just views?
- Are you grouping products by economics or just convenience?
- Are your audiences clear and logical or chaotic?
- Are you focusing Meta on the signal or just adding more noise?
This audit was a nice reminder for me that when a system fails, it’s not because of what you see on the surface – it’s because the surface is an imperfect reflection of a deeper, more fundamental issue. It’s tempting to try to fix the surface issue, but the real opportunity is digging deeper to figure out what’s really causing the results you’re seeing. I hope some of the ideas in this week’s issue help you do exactly that.
Talk to you in December!
All the best,
Sam
