You’re running 30 creative variants on Meta.
Different hooks. Different CTAs. Different background colours. You’ve tested sofa shots from the left versus the right. You’ve tried three different room settings.
Meta reports mediocre performance across all of them.
Here’s why: To Meta’s Andromeda system, they all look identical.
Not similar. Identical.
And when the platform can’t tell your creative apart, it can’t learn what actually works. You’re producing volume without variety. And it’s destroying your ability to find genuine winners.
Content
What Is Andromeda and Why Does It Matter?
Andromeda is Meta’s creative intelligence system. It’s the AI that processes your creative assets and determines how to serve them.
What Andromeda does:
- Analyses the visual and textual content of your ads
- Identifies patterns and similarities across your creative
- Groups similar creative together
- Learns which creative patterns drive performance
- Optimises delivery based on creative characteristics
The critical insight:
Andromeda doesn’t see your creative the way you see it. It processes images through computer vision, looking for visual patterns, compositional elements, colour palettes, and structural similarity.
What looks different to you often looks identical to Andromeda.
The Squint Test: How to See Creative Like Andromeda Does
Here’s the simplest way to understand if your creative has genuine variety:
Step 1: Open your Meta Ads Manager and navigate to your active campaigns
Step 2: Display 6-10 of your current creative assets side by side (use the grid view if available, or open multiple tabs)
Step 3: Squint your eyes until the images blur significantly
Step 4: Ask yourself: Can I still tell these apart?
If the answer is no, Andromeda can’t tell them apart either.
What happens when you squint:
- Text becomes illegible (just like it does to computer vision)
- Fine details disappear
- You’re left with composition, colour, and basic visual structure
- You see what the algorithm sees
This is how Andromeda processes your creative. And if they all look like the same blob of colour and shapes when you squint, the algorithm is treating them as the same creative with minor variations.
What Most D2C Brands Get Wrong About Creative Variety
Most brands think they’re creating variety when they’re actually just creating noise.
What false variety looks like:
Imagine a luxury sofa brand running these variants:
- Grey sofa in modern living room, shot from front-left angle, warm lighting
- Grey sofa in modern living room, shot from front-right angle, warm lighting
- Grey sofa in modern living room, shot straight-on, warm lighting
Different angles. Different copy. But fundamentally the same visual pattern: grey sofa, centre-framed, modern interior, warm tones.
When you blur your vision, these three images become nearly identical blobs. Same composition. Same colour dominance. Same context.
To you, they’re three different ads. To Andromeda, they’re the same ad with micro-variations.
What genuine variety looks like:
Now imagine these three variants instead:
- Close-up detail shot of sofa fabric texture, macro photography, cool morning light streaming through window
- Wide shot of sofa in evening setting, family gathered around, multiple people, warm ambient lighting, lived-in feel
- Minimal product shot, sofa isolated on white background, bright even lighting, architectural feel
When you squint at these, they remain visually distinct: different compositions, different lighting strategies, different colour dominance, different contexts.
This is variety Andromeda can actually learn from.
Why Creative Similarity Kills Performance
When Andromeda sees all your creative as the same, several things go wrong:
Problem 1: No Learning Signal
The platform needs distinct creative to understand what works. When everything looks the same, it can’t determine which creative characteristics drive performance.
You think you’re testing:
- Different room settings
- Different product angles
- Different copy approaches
The algorithm sees:
- Same visual composition
- Same creative pattern
- No meaningful differentiation
Result: The algorithm can’t learn, so it delivers everything semi-randomly.
Problem 2: Budget Fragmentation Without Benefit
You’re splitting budget across 30 variants that Andromeda treats as minor variations of the same creative.
Instead of:
- 3 genuinely distinct creative approaches with £5,000 each
- Clear learning about which approach works
You have:
- 30 similar variants with £500 each
- No clear signal about what’s working
- Perpetual learning phase
Problem 3: False Winners
When you have 30 similar variants, one will perform slightly better by chance. You scale it. Performance regresses.
This isn’t because the creative “fatigued.” It’s because it was never genuinely different in the first place. You were optimising for statistical noise within a single creative pattern.
Problem 4: Missed Opportunities
Whilst you’re testing 30 variations of “sofa in modern living room,” you’re missing entirely different creative approaches:
- Different contexts (outdoor terrace, cosy reading nook, busy family room)
- Different emotional tones (aspiration vs comfort vs sophistication)
- Different compositional structures (close-up details vs lifestyle scenes vs architectural shots)
- Different visual languages (warm and inviting vs cool and minimal vs rich and layered)
These might be where your actual winners live. But you’ll never know because you’re stuck in one creative pattern.
What Creates Genuine Variety for Andromeda
Andromeda recognises difference across several dimensions:
Dimension 1: Compositional Structure
Low variety:
- All centre-framed products
- All shot from same angle (front 3/4 view)
- All similar depth of field
High variety:
- Centre-framed vs off-centre vs environmental context
- Straight-on vs angled vs overhead vs detail shots
- Shallow focus vs deep focus vs architectural clarity
Dimension 2: Colour Palette Dominance
Low variety:
- All neutral tones (greys, beiges, whites)
- All warm lighting
- All similar saturation levels
High variety:
- Neutral vs rich/jewel tones vs monochrome
- Warm vs cool vs mixed lighting
- High saturation vs muted vs desaturated
Dimension 3: Visual Density
Low variety:
- All minimal, clean backgrounds
- All single-product focus
- All uncluttered compositions
High variety:
- Minimal vs layered vs busy environments
- Single product vs styled scene vs lived-in context
- Sparse vs abundant vs maximalist
Dimension 4: Context and Lifestyle Integration
Low variety:
- All empty room shots
- All aspirational, styled perfection
- All daytime lighting
High variety:
- Empty vs staged vs lived-in with people
- Aspirational vs relatable vs luxurious
- Morning vs afternoon vs evening vs night
Dimension 5: Scale and Perspective
Low variety:
- All full product shots
- All room-level perspective
- All similar framing
High variety:
- Full product vs detail close-ups vs environmental wide shots
- Eye-level vs low angle vs high angle
- Tight crop vs medium vs expansive framing
The key: You need variation across multiple dimensions, not just one.
Changing the copy whilst keeping everything else identical isn’t variety. Changing the room styling whilst keeping composition, lighting, and framing identical isn’t variety.
Genuine variety means fundamentally different creative approaches.
The Most Common Creative Similarity Traps
Trap 1: Template Addiction
You find a template that works. So you create 20 variants using the same template with different text overlays and product images.
Why it fails: Andromeda sees the same compositional structure across all variants. The template creates uniformity, not variety.
Trap 2: Same Photoshoot, Multiple Crops
You do one photoshoot in one location. You get 100 images from it. You create 30 ads using different shots from the same session.
Why it fails: Same lighting, same location, same styling, same time of day. Andromeda recognises this as the same creative session, regardless of which specific frame you use.
Trap 3: Room Variations
You think using different room settings creates variety. But they’re all:
- Modern minimalist interiors
- Neutral colour palettes
- Similar time of day
- Same architectural style
Why it fails: The context and aesthetic are identical. Andromeda sees one creative pattern in different rooms.
Trap 4: Colour Swap Variations
You create one room scene. Then you swap the sofa colour: grey version, blue version, beige version.
Why it fails: Identical composition, lighting, and context. Colour is just one dimension, and if everything else is the same, Andromeda treats these as minor variations.
Trap 5: Copy Testing Without Visual Change
You have one strong product image. You create 15 variants testing different headlines and CTAs.
Why it fails: The visual is 90% of what Andromeda analyses. Identical visuals with different text are treated as the same creative.
How to Audit Your Current Creative for Genuine Variety
Take your current active creative and run this audit using your own ads:
Step 1: The Squint Test (Visual)
Open your ads manager. Display all active variants. Squint until blurred. Count how many distinct visual patterns you actually see.
If you have 30 variants but only 2-3 distinct blurred patterns, you don’t have variety.
Step 2: The Composition Test
For each variant in your ads manager, identify:
- Subject placement (centre, left, right, environmental)
- Angle (straight-on, 3/4 view, overhead, detail)
- Depth (flat, shallow focus, deep focus)
Count how many different compositional approaches you have.
If 25 out of 30 are “centre-framed 3/4 view with shallow depth,” you don’t have variety.
Step 3: The Context Test
For each variant, identify the setting:
- Location type (living room, bedroom, outdoor, studio, etc.)
- Style (modern, traditional, eclectic, minimal, etc.)
- Occupancy (empty, staged, people present)
Count how many different contexts you’re showing.
If 90% are “modern living room, empty, aspirational,” you don’t have variety.
Step 4: The Colour Dominance Test
For each variant, identify the dominant colour palette:
- Warm vs cool
- Saturated vs muted
- Monochrome vs multicolour
Count how many different colour strategies you have.
If everything is “warm, muted, neutral palette,” you don’t have variety.
Step 5: The Lighting Test
For each variant, identify the lighting approach:
- Time of day (morning, afternoon, evening, night)
- Quality (soft, hard, mixed, natural, artificial)
- Direction (front-lit, side-lit, backlit, ambient)
Count how many different lighting approaches you’re using.
If everything is “afternoon, soft, front-lit,” you don’t have variety.
Target for genuine variety:
For every 10 creative variants, you should have:
- At least 4-5 distinct compositional approaches
- At least 3-4 different contexts or settings
- At least 3 different colour dominance strategies
- At least 3 different lighting approaches
If you’re not hitting these minimums, Andromeda sees your creative as repetitive, not varied.
Why This Matters More Now Than Before
Andromeda has become more sophisticated over the past two years. It’s better at identifying creative patterns and grouping similar creative together.
What this means:
Creative approaches that created “enough” variety in 2022 don’t create sufficient variety in 2025.
The bar for what counts as “genuinely distinct creative” has risen.
Brands still using 2022 creative variety strategies are seeing:
- Longer learning phases
- Less efficient delivery
- Weaker performance from “variant testing”
- Higher CPMs as the platform struggles to differentiate
This is why many D2C brands feel like “Meta performance got worse.” Meta didn’t get worse. Your creative variety approach became obsolete.
The Connection to Micro-Moment Targeting
Here’s where this connects to strategic performance marketing:
When you build creative for specific micro-moments, you naturally create genuine variety.
Thursday 6pm “weekend preparation” moment for luxury furniture:
- Context: Evening at home, planning weekend entertaining
- Composition: Lifestyle scene, people present or implied, warm and inviting
- Colour: Rich, warm tones, ambient evening lighting
- Copy: Weekend entertaining, preparation, hosting
Sunday 2pm “nest building” moment:
- Context: Afternoon, relaxed, personal sanctuary feeling
- Composition: Close-up comfort details, tactile textures, intimate framing
- Colour: Soft, muted, cocooning palette
- Copy: Comfort, retreat, personal sanctuary
Tuesday 10am “investment decision” moment:
- Context: Clear daytime, rational evaluation setting
- Composition: Architectural, clean, product-focused, objective
- Colour: Bright, neutral, clarity-focused
- Copy: Quality, craftsmanship, investment value
These are fundamentally different creative approaches because they’re designed for fundamentally different moments and motivations.
What to Do If Your Creative Lacks Variety
If you’ve done the squint test on your own ads and realised they all look the same, here’s what to do:
Immediate Action: Consolidate
Stop running 30 similar variants. Consolidate to your 3-5 most visually distinct performers and give them proper budget.
Better to run 5 variants with £3,000 each than 30 variants with £500 each when they’re all visually similar anyway.
Short-term Fix: Diversify Existing
From your current creative pool, identify the most visually distinct variants:
- Different compositions
- Different contexts
- Different colour palettes
- Different lighting approaches
Keep only the ones that remain distinct when you squint. Pause the rest.
Medium-term Solution: Produce Strategic Variety
When producing new creative, mandate variety across multiple dimensions:
For your next 6 creative variants:
- 2 must be product-focused architectural shots, 2 lifestyle scenes, 2 detail close-ups
- 2 must be warm evening tones, 2 cool morning light, 2 bright midday
- 2 must be minimal/empty, 2 styled, 2 lived-in with people
- 2 must be centre-framed, 2 environmental wide, 2 tight detail crops
This forces genuine variety.
Long-term Strategy: Build for Moments
Implement micro-moment targeting where each moment naturally requires different creative approaches.
Different moments = different contexts = different compositions = genuine variety.
This creates variety as a byproduct of strategy, not as a forced creative exercise.
The Bottom Line on Creative Variety
Most D2C brands are running 20-40 creative variants that all look identical to Meta’s Andromeda system.
They think they’re testing. They’re actually just fragmenting budget across minor variations of the same creative pattern.
The squint test reveals this immediately. Open your ads manager, squint at your active creative, and see how many truly distinct patterns emerge. If you can’t tell them apart when you squint, Andromeda can’t either.
And when the platform can’t differentiate your creative, it can’t learn what works. You’re stuck in perpetual learning phase, scaling false positives, and missing genuine opportunities.
Genuine variety isn’t about volume. It’s about creating fundamentally different creative approaches across composition, context, colour, lighting, and perspective.
This is how you give Andromeda the signal it needs to actually learn and optimise.
The Graygency helps D2C brands grow profitably by identifying high-propensity buying moments using third-party data, creating targeted creative for those moments, and building growth systems that compound over time.
Your creative all looks the same to Meta. Let’s fix that.
