See what makes your best ads work

Every ad image in your account is automatically scanned across 40+ dimensions. No manual tagging. No spreadsheet tracking. Nylo's computer vision (AWS Rekognition + AI classification) reads what's actually in your creative.

What gets detected:

• Hook type. Question, bold claim, curiosity gap, social proof, before/after. Which hooks convert?
• Emotion. Happy, excited, calm, urgent, surprised. What emotional tone drives the best ROAS?
• Talent. Influencer, founder, customer, actor, none. Does featuring a face improve performance?
• Visual style. Lifestyle, product shot, UGC, graphic design, flat lay. What aesthetic works for your audience?
• CTA presence. Is there a visible CTA? Does it impact click-through?
• Brand elements. Logo visible? Product in frame? Text overlay content (OCR)?
• Color and composition. Dominant color, color palette, object detection, brand safety score.

A DTC beauty brand ran creative analysis across 340 Meta ad images and discovered that UGC-style creatives with a "curiosity gap" hook and no visible logo outperformed studio product shots by 62% on ROAS. The creative team doubled down on that format and cut their cost per acquisition by 28%.

Videos are analyzed frame by frame. Not just "what's in the video" but when things happen and how the structure affects performance.

What gets analyzed:

• Hook quality (first 3 seconds). Content, emotion, text presence, face presence. The first 3 seconds determine whether someone watches or scrolls.
• Emotional arc. Frame-by-frame emotion progression with confidence scores. Does the ad build tension? Resolve it? Shift emotions?
• Product introduction timing. Exact second the product first appears. Is it immediate (under 1s), early (under 3s), mid (under 10s), or late? How does timing correlate with completion rate?
• Scene structure. Number of scenes, transitions, pacing. Faster cuts or longer holds?
• End card. Has CTA, has logo, end card text content. Does a strong end card improve conversion?

Each video is broken into keyframes (0s, 1s, 3s, 5s, 10s, 15s) and each frame is analyzed independently, then combined into a structural narrative.

The real power isn't in analyzing one creative. It's in finding patterns across hundreds.

Nylo correlates creative attributes with performance metrics (ROAS, CPA, CTR, completion rate) to surface patterns like:

• "Creatives with a question hook and a face in the first frame have 2.3x higher CTR"
• "Videos that introduce the product before 3 seconds have 40% higher completion rate"
• "Warm color palettes outperform cool tones on weekend audiences by 18%"

These patterns are discovered from your data, not industry benchmarks. What works for your audience, your products, your campaigns.

You're not locked into predefined categories. Define your own creative classification system:

• Custom hook types for your industry
• Custom visual style categories
• Custom emotion labels
• Custom audience segment tags

The AI classification adapts to your taxonomy, so the insights map directly to how your team thinks about creative strategy.

Stop guessing which creative direction to pitch. Show clients data on what visual styles, hooks, and formats actually convert for their audience. Turn "creative review" from a subjective meeting into a data-driven conversation.

Scale creative production by knowing what works before you spend. Test fewer variants with higher confidence. Identify winning patterns and brief your creative team with evidence, not opinions.