One ecommerce brand went from $54,000 in weekly Meta spend to $622,000 in the same period, year over year. Not by finding a single winning ad. Not by cranking up budgets on what already worked. By obsessing over creative diversity as a system for feeding Meta’s algorithm the signals it needs to find new buyers at scale.
This isn’t a theory. It’s a playbook that’s emerging from how Meta’s Andromeda retrieval engine actually works, backed by internal Meta workshops and in-platform data most advertisers haven’t seen yet.
Here’s how creative diversity became the single biggest scaling lever on Meta in 2025, and why the brands investing in it now are creating a compounding advantage that will be hard to catch.
Meta’s Andromeda changed the game: creative is targeting now
If you’re still thinking about Meta Ads through the lens of audiences and interest targeting, you’re operating on outdated logic.
Meta’s Andromeda update replaced the old deterministic model (if user did X, show them Y) with a probabilistic one. The algorithm now matches ads to users based on relevancy scores generated from creative signals: engagement, watch time, hook rate, and conversion patterns.
In practical terms: your creative is your targeting. Every ad you put into the system sends a signal about who it resonates with. Meta latches onto that signal and finds similar people. The more diverse and differentiated your creative signals, the more distinct audience pools Meta can tap into.
This is why the old approach of iterating on the same static with different copy no longer moves the needle. You’re sending the same signal to the same audience, over and over.
The creative similarity problem most brands don’t know they have
Meta recently ran internal workshops with select agencies where they scored live ad accounts for creative similarity: a metric that measures how much Andromeda perceives two ads as functionally identical.
The gap was stark.
Two UGC creators, different people, both filming in a kitchen, delivering a similar message. Meta’s system scored them as the same asset. Two completely different individuals, but because the setting, format, and message were similar, Andromeda grouped them together and served them to overlapping audiences.
The investment in producing that second creative was largely wasted. It wasn’t driving incremental reach. It wasn’t tapping a new audience pool. It was competing with the first ad for the same eyeballs.
Meta has codified this into a sliding scale of differentiation:
- Left (low differentiation): Similar message + similar visual style. Same angle, same format, similar creator, similar hook. These ads cannibalize each other.
- Center: Different message, same style. You’re testing a new script angle, but the format is identical. Some incremental value, but limited.
- Right (high differentiation): Different message + different visual style. New angle, new person, new vehicle, new visual journey. This is where incremental reach compounds.
The brands that live on the right side of that scale are the ones 10x’ing spend. The ones stuck on the left wonder why their CPA keeps climbing despite “testing new creative.”
The persona pyramid: a framework for infinite creative diversity
The question most brands ask at this point is: “How do we come up with enough differentiated creative to sustain this?”
The answer is a layered approach that starts from the bottom of what we call the persona pyramid and multiplies upward:
Layer 1: Product. Your product is the foundation. For most ecommerce brands, you have multiple products or SKUs, each with distinct benefits.
Layer 2: Personas. For each product, how many distinct buyer personas exist? If your TAM is broad, you might identify 10 to 12 personas with fundamentally different motivations, pain points, and emotional triggers.
Layer 3: Angles. Within each persona, there are multiple angles you can take. Different pain points, different desired outcomes, different emotional narratives. Optimistically, 15 to 20 angles per persona.
Layer 4: Vehicles. How you deliver the message: podcast ad, VSSL, founder story, supermarket walkthrough, UGC testimonial, split-screen comparison, first-person POV. Each vehicle appeals to different content consumption preferences.
Layer 5: Visual style. The setting, the framing, the color palette, the pacing. Two ads with the same script but completely different visual journeys will reach different people.
Do the math. 3 products x 12 personas x 20 angles x 5 vehicles x 3 visual styles = 10,800 unique concepts. You’re not running out of ideas. You’re under-investing in the systematic exploration of what’s available.
The priority isn’t to produce all 10,000. It’s to understand that the creative well is effectively infinite, and your job is to prioritize the highest-impact combinations while maintaining a consistent pipeline of genuinely differentiated assets.
The KPIs nobody is tracking (but should be)
Most ad accounts track ROAS, CPA, and maybe CTR. Almost nobody tracks the metrics that actually predict whether your creative strategy is working at the Andromeda level.
1. Incremental reach volume
Request incremental reach reports from your Meta rep monthly. This tells you how many net-new accounts (not impressions, accounts) your ads reached that hadn’t seen your brand before.
As you scale spend, incremental reach as a percentage will naturally decrease. That’s fine. What matters is that the volume keeps increasing. If volume plateaus while spend goes up, you’re saturating your audience pool and creative diversity is the fix.
A healthy range: 10 to 20% incremental reach relative to brand size. Below 10%, you’re hitting diminishing returns fast.
2. CPMR (cost per thousand accounts reached)
This metric is available directly in Meta Ads Manager. It’s essentially CPM multiplied by frequency, giving you the true cost of reaching unique accounts rather than just serving impressions.
Add it to your column mix today. Watch it on a weekly trend. You’ll notice it spikes downward after every creative sprint where you inject genuinely diverse assets, then levels out as those assets saturate their audience pools.
The red flag: CPMR increasing proportionally (or faster) than your spend increase. That means you’re paying more to reach the same people.
On an asset level, CPMR reveals even more. When you find a strong unaware or problem-aware creative unlock, its CPM can be 5 to 6x lower than your account average. That’s the signal that it’s reaching a fundamentally different audience pool and dragging performance up across the entire account.
3. New visitor percentage
Track this both at the account level and by asset format. It tells you not just whether your creative is getting new eyeballs (that’s CPMR), but whether those new eyeballs are actually clicking through to site.
The pattern is consistent: video ads with new angles and personas drive higher new visitor percentages. Statics and iterations on existing concepts tend to have lower new visitor rates; they’re better at sweeping up the audience that video brought in.
There’s a strong correlation between declining new visitor percentage and declining performance. New visitor % is a lagging indicator of rising CPMR, which itself is a leading indicator of declining CPA performance. By the time your CPA spikes, the creative diversity problem has been building for weeks.
What “differentiated” actually means in practice
Knowing you need differentiation is one thing. Executing it is another. Here are the levers that actually move the similarity score, ranked by impact:
Visual journey differentiation (highest impact)
The full visual experience the user goes through in the first 3 to 5 seconds is the strongest differentiator. Not just “different creator,” but a completely different setting, camera angle, pacing, and visual flow.
Going from sitting in a car to first-person POV to a supermarket walkthrough to a studio setup: each of these sends a fundamentally different signal to Andromeda, even with a similar core message.
Dr. Squatch does this well: their hook variations are visually unrecognizable from each other, but the body of the video can be similar. The opening visual journey is what creates the differentiation.
Persona and angle diversification (high impact)
Different motivators unlock new audiences 89% of the time, according to Meta’s internal data. This comes down to finding new angles rooted in different personas.
One powerful tactic: reframing the same benefit through different psychological motivations. Someone predisposed to loss aversion responds to “stop wasting money on X” while someone motivated by gain responds to “unlock Y.” Same product benefit, completely different emotional entry point.
Another example: framing the target audience differently within the creative. An ad where a 25-year-old talks about a product for his mom hits a different audience than an ad featuring someone over 45 using the same product directly. Same message, different framing, different delivery.
Vehicle diversification (medium impact, quick win)
This is often the fastest path to creative diversity. Take an existing winning script and reproduce it as a podcast ad, a VSSL, a founder story, a TikTok-native format, or an influencer story.
Different vehicles appeal to different content consumption preferences. Someone who converts through a podcast-style ad might never engage with traditional UGC. You’re essentially unlocking additional scale from angles you’ve already validated.
TikTok’s organic formats are your best testing ground for vehicle inspiration. The formats that perform organically tend to translate well into paid creative, and they’re usually cheaper to produce than high-production alternatives.
Creator diversification alone (lower impact than before)
Six months ago, simply using different creators was a major lever. That’s shifted. Unless your creators represent genuinely different demographics (different genders, age ranges, backgrounds) that naturally pair with different motivations and pain points, swapping creators without changing the visual journey or angle produces diminishing returns.
Meta is getting better at recognizing when two creatives are functionally identical regardless of who’s on camera. Creator diversity works best when it’s a byproduct of persona and angle diversity, not the primary differentiator.
Scaling tactics that compound creative diversity
Creative diversity is the strategic foundation. But there are tactical execution details that determine whether your scaling actually holds. Here are the moves that separate brands spending $100K/month from those spending $600K+.
Horizontal scaling with manual bid constraints
The classic vertical scaling approach (raise budget, pray) breaks past a certain spend level. Horizontal scaling, launching multiple campaigns targeting the same funnel stage with manual bid constraints, is more reliable.
The timing matters: launch new horizontal campaigns before weekends and high-traffic periods. Meta’s delivery system needs time to calibrate, and launching into a high-volume window gives the algorithm more data points to optimize against in the first 48 hours.
Put previous winners into bid cap with new ASC
When you find a creative that’s proven but starting to fatigue in broad targeting, move it into a bid-capped Advantage Shopping Campaign. The new ASC algorithm handles bid cap mechanics better than it did even six months ago.
This creates a controlled environment where your winners can extract more value from specific audience segments at a capped cost per acquisition. It extends the useful life of strong creative while your team produces the next wave of diversified assets.
Don’t turn off ads in scaling campaigns
This is counterintuitive for anyone trained on the “kill underperformers fast” playbook. In non-testing campaigns that are actively scaling, turning off individual ads disrupts Meta’s delivery optimization in ways that hurt overall campaign performance.
The exception is testing campaigns, where the goal is rapid signal collection. There, aggressive pruning makes sense. But in your scaling campaigns, let the algorithm redistribute spend naturally. An ad with a higher CPA today might be the one reaching the incremental audience that feeds your retargeting funnel tomorrow.
Plan for longer learning periods
Meta’s learning periods are getting longer, especially for new campaigns at higher spend levels. The algorithm needs more consolidation at the adset level before it finds optimal delivery patterns.
The practical implication: budget your first 7 to 10 days of a new campaign as learning investment, not performance expectation. If you’re judging a new campaign on Day 3 metrics and killing it, you’re wasting the learning data Meta already collected and starting the cycle over.
Consolidation at the adset level matters more than ever. Fewer adsets with higher budgets learn faster than many adsets with fragmented spend.
High production content is growing as a percentage of spend
This one cuts against the “UGC is king” narrative. The data shows high-production content, think cinematic brand films, polished product demonstrations, studio-lit lifestyle shoots, is growing as a percentage of total ad spend among top spenders.
It’s not replacing UGC. It’s complementing it. High-production assets tend to perform well with audiences that don’t respond to typical UGC formats, which is exactly the creative diversity thesis. The combination of polished and raw, professional and authentic, hits more audience pools than either alone.
If you’re exclusively running UGC, you’re leaving an entire creative vehicle (and its associated audience) on the table.
Three new Meta metrics coming to your ad account
Meta is rolling out three new in-platform metrics that will make creative diversity measurable without needing rep-only reports:
Creative fatigue score
A percentage score on each asset showing how saturated your target audience is becoming with that specific creative. When interaction rates decline due to overexposure, this score climbs.
Previously, you could feel fatigue through CPA increases, but you lacked a leading indicator. Now you’ll see it coming. If a winning asset hits 30 to 40% fatigue, that’s your signal to have replacements ready.
The brands in the strongest position here are those with 30 to 50+ high performers in their ASC at any given time. When one fatigues, spend redistributes naturally. If you only have 3 performers and they’re all fatiguing simultaneously, you’re in danger zone.
Creative similarity score
This brings the workshop analysis into self-serve. You’ll see directly when Andromeda considers two of your ads functionally identical, letting you identify wasted creative investment in real time.
Expect this to be revealing. Assets you assumed were “different enough” will likely score as highly similar. The threshold for true differentiation is higher than most advertisers realize.
Top creative themes
A breakdown of active ads and spend across creative themes: humor, nostalgia, savings/offers, social proof, and more. This helps you spot gaps in your creative mix and identify where you’re over-investing.
If 60% of your spend is on social proof testimonial ads and 0% on humor or satire, you’ve found an obvious diversification opportunity. Satire and humor consistently overindex in top performers, yet many brands never test them.
Partnership ads: a quick CPMR win
One tactic worth calling out: partnership ads (formerly branded content ads) using native Instagram story-style statics.
The format: a static that looks exactly like an organic Instagram story from an influencer or creator, paired with the partnership ad label. The combination drives remarkably cheap CPMs and high click-through rates because it looks native, not like an ad.
The result: higher relevancy scores from Meta, which translates to better conversion rates across partnership ad campaigns. We’re seeing consistent CPA improvements when this format is deployed at the lower funnel.
The impact on incremental reach is less dramatic than persona-diversified video. Our hypothesis is that partnership ads excel at converting the more “stubborn” audience segments who weren’t converting through standard formats, rather than reaching entirely new pools. But for squeezing more juice from existing audiences, it’s a high-ROI play.
The brand identity tension (and how to resolve it)
There’s an uncomfortable question in all of this: what happens to brand consistency when the algorithm rewards visual chaos?
Look at AG1’s ad library 18 months ago: every asset had a consistent brand look and feel. You could identify the brand without seeing the logo. That consistency built brand equity, but it also meant Andromeda was grouping most of their ads together and limiting incremental reach.
Their more recent library tells a different story: much more color diversity and visual variation. They adapted.
The resolution isn’t “abandon your brand.” It’s redefining what brand means in this context. Brand isn’t the color palette or the visual template. It’s the feeling you create in the customer’s mind. You can be wildly diverse in visual execution while maintaining a consistent value proposition, tone, and emotional signature.
Brands that are more protective of visual consistency will find Meta scaling harder. This is a structural reality of how Andromeda works, not a preference. The sooner brand teams internalize this, the sooner they can find the balance between identity and diversity that lets them scale.
Where AI accelerates the flywheel
The volume requirements of this creative diversity approach are significant. Some brands need hundreds of genuinely differentiated ads per month. That’s expensive and slow with traditional production.
AI video generation tools like Sora 2 are compressing timelines and costs meaningfully. The quality improvements are substantial: better realism in UGC-style content, more natural camera movement, and dramatically faster iteration cycles.
The key capability for creative diversity specifically: remixing. Taking a real product, a real person’s likeness, and placing them in completely different visual contexts at a fraction of the production cost. That’s exactly what Andromeda rewards: the same core message in visually distinct environments.
This isn’t about replacing human creative strategy with AI slop. It’s about using AI to execute on the visual diversification layer at the speed and scale that the algorithm demands, while human strategists focus on the persona, angle, and messaging layers where strategic thinking matters most.
Key takeaways
Ready to scale your Meta Ads with a creative diversity framework?
At Naniza, we build systematic creative testing pipelines that feed Meta’s algorithm the diverse signals it needs to find your next audience pool. Our clients consistently see compounding returns as creative volume and diversity increase over time.
-720x403.png&w=3840&q=75)
-720x403.png&w=3840&q=75)