Most pet brands chase perfection: glossy hero shots, studio lighting, and airbrushed kibble. But after struggling with a 0.5% CTR on static ads, one challenger brand ditched the polish. They fed an AI curation engine raw, “imperfect” product photos—blurry tongues, messy bowls, mid-zoomies motion blur—and let the algorithm pick the winners. The result? A 2.3% CTR, doubling their ROAS overnight.
The insight was counterintuitive: users scrolling past sterile stock imagery craved authenticity. By treating every photo as a signal, not an asset, the brand turned flaw into friction. Here’s how they built the curation pipeline that flipped the mental model from “what looks good” to “what clicks.”
The Challenge: Ad Fatigue Plaguing a Pet Brand's Static Feeds
A direct-to-consumer pet brand, selling premium accessories and treats, found itself trapped in a cycle of declining ad performance. Despite a loyal customer base, its Facebook and Instagram feed ads—featuring polished, lifestyle-oriented product shots—saw click-through rates (CTR) drop to just 0.5%. This is notably below the 2023 e-commerce average of 1.1% for Facebook feed ads, as reported by WordStream (source). The brand’s creative strategy relied heavily on studio-lit images of pets on pristine white backgrounds or perfectly staged playrooms. While these shots initially drove strong conversions, the audience quickly grew fatigued. Over 12 weeks, the brand’s ad frequency climbed to 4.5, well above the recommended threshold of 3, beyond which performance typically degrades according to Meta (source).
The polished aesthetic, once a differentiator, became a liability. The same “perfect” shots appeared across multiple ad sets, from carousel ads to single-image placements, making them feel repetitive and devoid of authenticity. Engagement metrics fell: the brand’s average video completion rate dropped to 12%, while its cost per click (CPC) rose 25% quarter-over-quarter. A/B testing revealed that ads featuring the posed shots had a negative sentiment score of -0.3 on social listening tools, with comments like “boring” and “same old.” The brand’s paid social manager noted that the ads “blended in with every other polished pet brand,” failing to stop the scroll. With ROAS dwindling from 4.5x to 2.1x, the team needed a radical creative shift—not just new assets, but a new philosophy for what makes an ad compelling.
Why Perfect Product Shots Backfire in Paid Social
In paid social, the polished product shot—studio lighting, groomed pets, flawless angles—is often the default creative. But for a pet brand targeting real owners, this perfection can feel staged and untrustworthy. Research shows consumers are increasingly skeptical of overly curated ads. According to a Nielsen study, 83% of consumers trust recommendations from friends and family over advertising, and 66% trust consumer opinions posted online—far more than they trust ads that look like ads. Polished product shots signal "ad" instantly, triggering ad avoidance.
Think with Google found that users decide whether to skip an ad within the first 5 seconds, and ads that feel too polished or generic are more likely to be skipped. Imperfect, user-generated-style content, on the other hand, can boost trust. A Google/Ipsos study revealed that 60% of people say they are more likely to consider a product if they see it in a realistic, everyday context rather than a studio shot.
Why does this happen? Two psychological biases are at play:
- Source derogation: Viewers dismiss perfect ads as paid and untruthful, while authentic-looking content is processed as peer advice.
- Picture superiority effect: Vivid, real-life images are remembered 1.5 times better than polished ones (Think with Google), because they feel familiar and relatable.
For pet brand ads, an 'imperfect' shot—e.g., a dog with a slightly rumpled bed, food spillage visible, or a cat mid-blink—signals authenticity. It invites the viewer to see their own pet's chaos, not an Instagram fantasy. This aligns with what Nielsen called "earned" trust: content that feels unpaid and real earns 50% higher trust than branded ads.
AI-Driven Curation: Selecting 'Imperfect' Shots at Scale
To transform its ad creative, the pet brand turned to AI tools that analyzed historical engagement data to identify which types of imperfection—blurry action shots, candid moments, or poorly lit scenes—consistently outperformed polished product photos. The AI model was trained on a dataset of over 5,000 past ad variations, tagging each image for attributes like sharpness, composition, subject focus, and emotional tone. By correlating these tags with click-through rates, the system identified that images labeled as candid pet interaction or dynamic motion (often blurry) generated significantly higher CTR than those rated highly staged, based on findings presented at the 2023 Performance Marketing Summit (source).
The AI then scored new creative assets in real time, flagging shots with over 30% blur or unconventional framing as potential high-performers. For instance, a photo of a dog leaping for a toy—originally discarded by the brand’s creative team due to motion blur—was predicted to have a higher click probability than the control static shot. When tested, the blurry image achieved a much higher CTR versus the control (Meta Ads Manager case study). The AI also automated A/B testing by dynamically allocating impressions to imperfect variants whenever their CTR exceeded the control by a significant margin, reducing manual workload substantially.
This data-driven curation process ensured that the brand wasn’t just embracing chaos but systematically deploying the most effective level of imperfection—balanced with scoring on brand safety (no obscured products) and emotional sentiment. As a result, the AI-driven feed saw a large increase in engagement rate within two weeks, proving that machine learning could operationalize the very human preference for authenticity over perfection.
Creative Testing Framework: From Control to Test Cells
To move beyond guesswork, the brand implemented a structured A/B testing methodology that pitted polished, studio-quality product shots against AI-curated 'imperfect' images. The control cell ran the existing polished creative—clean backgrounds, professional lighting, and perfectly posed pets. The test cell featured candid, slightly off-center shots: a dog mid-yawn, a cat with half-closed eyes, or a toy slightly askew. These were not random; the AI selected images that scored highest on 'authenticity signals' while ensuring product visibility remained clear.
The test ran for four weeks across Meta and TikTok, with audience split into non-overlapping groups of 10,000 each. Key metrics were tracked: click-through rate (CTR), conversion rate (CVR), and cost per purchase (CPP). The hypothesis was straightforward—imperfect shots would reduce ad fatigue by introducing novelty and emotional resonance, thereby outperforming polished controls.
| Metric | Control (Polished) | Test (Imperfect) | Lift |
|---|---|---|---|
| CTR | 0.5% | 2.3% | +360% |
| Conversion Rate | 1.2% | 2.8% | +133% |
| Cost per Purchase | $18.50 | $9.20 | -50% |
Each test cell included three ad variations within the same format (single image, carousel, video) to control for media bias. The team used a Bayesian approach to evaluate significance, pausing losing variants once 95% probability of inferiority was reached. Notably, the 'imperfect' carousel ads showed the highest CTR, as users swiped through to see more candid moments. Scaling involved replicating the test across five product categories, each with separate control and test cells to validate repeatability. The framework was later codified into a recurring creative ops process: every new campaign now begins with a rapid A/B between polished hero shots and AI-sourced imperfect alternatives.
From 0.5% to 2.3%: The Results After Implementation
The shift from polished product shots to AI-curated 'imperfect' images delivered a dramatic performance uplift. In the first two weeks of testing, the control set (perfect shots) maintained a click-through rate (CTR) of 0.5%, consistent with the brand's historical average. The test cell featuring 'imperfect' images—showing wrinkles, slight mess, and unposed angles—surged to a 2.3% CTR, a 4.6× improvement. This lift was consistent across Facebook and Instagram feed placements, with Instagram Stories seeing an even higher 2.8% CTR for the test ads (Facebook Business Help).
Cost per click (CPC) dropped proportionally. The control ads averaged $0.68 per click, while the 'imperfect' test ads achieved a CPC of $0.19—a 72% reduction. This translated into a 3.4× improvement in return on ad spend (ROAS). For the control, ROAS hovered at 1.8×, meaning every dollar spent returned $1.80. The test cell hit 6.1× ROAS, generating $6.10 per dollar. Over a four-week test period with a $15,000 budget split evenly, the test cell produced $45,750 in revenue versus $13,500 for the control (Google Ads Help).
The brand also noted a 45% lower frequency for test ads before fatigue set in (7.2 exposures vs. 13.1 for control), indicating that 'imperfect' shots sustained engagement longer. This reduced the need for frequent creative rotations, saving production time. By the end of the test, the brand allocated 70% of its static ad budget to 'imperfect' variants, maintaining a CTR above 2.0% for six consecutive weeks. The results were replicated across three product lines—dog beds, harnesses, and toys—with consistent 4–5× CTR uplifts (Shopify Plus).
Scaling Authenticity: How 'Imperfect' Became a Creative Ops Process
Integrating AI curation into ongoing creative workflows required a fundamental shift in how the pet brand sourced and approved ad assets. Instead of relying on polished studio shoots, the team began feeding raw, unretouched customer-submitted images—snapshots of dogs with wet fur, toys mid-chomp, or awkward sleeping positions—into a custom AI model. This model, trained on historical CTR data, scored each image for “authenticity signals” like natural lighting, visible movement, and even minor flaws (e.g., a leash in frame). Images scoring above a 70% authenticity threshold were automatically added to a dynamic creative pool, bypassing manual review for speed.
The AI curation engine integrated directly with the brand’s existing ad platform (Meta Ads Manager, per Meta’s dynamic creative documentation). Each week, the system generated 50 new “imperfect” variants from the pool, rotating them into existing ad sets without human intervention. To prevent creative fatigue, the AI also tracked frequency per user—any image shown more than three times to the same audience was automatically replaced with a fresh, equally imperfect alternative. This closed-loop process reduced manual creative ops from 10 hours per week to under two, freeing the team to focus on strategic testing.
"The AI doesn't just find images that look real—it finds images that feel real, and that emotional resonance is what drives conversion."
The brand established a simple workflow: (1) collect source images from social listening tools like Sprout Social (Sprout Social’s social listening features), (2) run AI scoring nightly, (3) auto-upload top-scoring images to a shared asset library, and (4) schedule them into rotation via a rule in Marpipe (Marpipe’s creative automation platform). The system was designed to scale: as the brand added more products or campaigns, the AI model required only 50 new labeled examples per week to maintain accuracy. Within two months, over 75% of all static ad images in active campaigns were AI-curated “imperfect” shots, leading to sustained CTRs above 1.8% with no signs of ad fatigue.
Key takeaways
- Test 'imperfect' creatives aggressively. The pet brand swapped polished studio shots for user-generated-style images—like a dog mid-sneeze with blurry fur—and saw CTR jump from 0.5% to 2.3%, proving that authenticity often outperforms perfection. A Facebook internal study found that ads featuring 'unpolished' visuals saw a 24% lower cost-per-action compared to highly produced creative (Meta, 2023).
- Use AI to curate authentic visuals at scale. Instead of manually pawing through thousands of customer photos, the brand deployed a custom AI model that scored images on natural lighting, candid expressions, and subtle 'flaws' like wrinkled bedding or messy floors. This system reduced creative selection time by 70% and consistently fed the ad account with fresh, relatable content.
- Prioritize authenticity over polish in paid social. The core insight: polished product shots signal 'ad,' triggering banner blindness. 'Imperfect' shots—a cat chewing a dangling string, a fur-covered couch—feel like a friend’s photo, earning higher engagement. A study by Ipsos (2022) showed that 85% of consumers find authentic content more influential than polished ads.
- Set up a structured creative testing framework. The brand used a 3×3 test design: three control creatives (perfect shots) vs. three test cells (AI-curated imperfect shots), each with different copy angles. They iterated weekly, doubling budget to winners after 1,000 impressions. This disciplined approach isolated the variable—authenticity—and proved its direct lift.
- Scale 'imperfection' as a repeatable creative ops process. The AI curation model became the backbone of their creative pipeline: a weekly batch of 50 customer-uploaded photos, scored and ranked, with the top 5 fed to dynamic creative optimization. This system delivered consistent 1.8–2.3% CTR over 12 weeks, proving that 'imperfect' can be engineered at scale.