---
name: visual-benchmarking
description: "Benchmark, compare, and improve competitive analysis pages by auditing visual richness across 6 categories, scoring against gold-standard references, and prescribing specific asset placements. Use when asked to: compare brand pages, improve visual quality, audit image richness, benchmark against competitors, run a visual critique, or improve a competition page. Triggers on: 'benchmark', 'compare pages', 'visual critique', 'improve the page', 'make it richer', 'comparison and improvement', 'visual audit'."
---

# Visual Benchmarking & Improvement

Score, compare, and improve competitive analysis pages by auditing visual asset coverage across 6 mandatory categories.

## 6-Category Visual Richness Benchmark

Every brand page must cover ALL 6 categories with 3+ assets each to score 4/5:

| # | Category | What qualifies | Reliable sources |
|---|----------|---------------|-----------------|
| 1 | Product UI | Real app screens showing features in use | App Store, Google Play, Mobbin |
| 2 | Campaign/OOH | Paid media, billboards, digital ads | Brand social accounts, agency sites |
| 3 | Lifestyle | Real people using product in context | Brand Instagram, press kits, partnership imagery |
| 4 | Brand system | Design tokens, typography, color in application | Live site CSS extraction, brand guidelines pages |
| 5 | Agency/editorial | Case studies, press, design community coverage | Behance, Dribbble, Brand New, agency portfolios |
| 6 | Physical/merch | Cards, packaging, physical brand objects | Product photography, unboxing content |

## T1/T2/T3 Image Quality Tiers

Within each category, every image gets a quality tier. T1 is the only tier that creates visual impact.

| Tier | Name | Definition | Rule |
|------|------|-----------|------|
| **T1** | Editorial | Campaign photography, lifestyle, OOH, agency creative, people-in-context | Min 8 per page. Hero must always be T1. |
| **T2** | Product | Card renders, hero UI in context, feature screenshots, app mockups | Max 40% of total image count |
| **T3** | Reference | Icon grids, color swatches, typography specimens, small UI crops | Max 20% of total. Never use as filler. |

**Key insight:** Image count is the wrong metric. Adding T3 icons pads the count while actively lowering visual quality. v3 with 49 images scored worse than v2 with 20, because T3% was 33% and T1% was only 8%.

## Scoring Dimensions (1–5 each)

| Dimension | What to evaluate |
|---|---|
| Product clarity | Can you understand the product from visuals alone? |
| Visual richness | Depth beyond basic screenshots — variety of section types and image categories |
| Emotional pull | Does it make you want to use the product? |
| Information density | Right amount of visual content per section |

## Audit Process

### Step 1: Count and classify
- Count total `<img>` tags in HTML
- List all asset files in the brand's folder
- Classify each as: used/unused, category (1–6), T1/T2/T3 tier
- Flag junk: 404 pages, CAPTCHA captures, blank/corrupt files, exact duplicates

### Step 2: Score current state
- Score each of the 4 dimensions (1–5)
- Count categories covered (need 3+ assets per category)
- Record: X/6 categories, T1/T2/T3 percentages

### Step 3: Identify gaps
- Which categories have <3 assets?
- Which unused assets should be on the page?
- What new assets need sourcing?
- Is T1% above 20%? Is T3% below 20%?

### Step 4: Write critique.md
Post content as text in the comms thread (agents can't write to server). Kitt commits.

```
# [Brand] — Visual Audit & Critique
Auditor: [name]
Date: [date]
Target: [brand]-v[N]/
Status: RESEARCH_COMPLETE

## Scores
| Dimension | Score | Notes |
|-----------|-------|-------|
| Product clarity | X/5 | ... |
| Visual richness | X/5 | ... |
| Emotional pull | X/5 | ... |
| Info density | X/5 | ... |

## Category Coverage: X/6
| # | Category | Assets | Status |
|---|----------|--------|--------|
| 1 | Product UI | X | ✅/⚠️/❌ |
...

## T1/T2/T3 Breakdown
T1: X (X%) | T2: X (X%) | T3: X (X%)

## High-value unused assets
| Filename | Category | Tier | Recommended placement |
|----------|----------|------|-----------------------|
| ... | ... | ... | ... |

## Junk files (delete)
- filename.jpg — reason

## Compression flags (over limits)
- filename.jpg — Xkb (limit 200KB JPG / 300KB PNG)

## New assets needed
| Category | Description | Source |
|----------|-------------|--------|
| T1 Lifestyle | Real people with product | Revolut Instagram |
...

RESEARCH_COMPLETE ✅
```

## Quality Gate (before build starts)
- [ ] T1 images ≥ 8
- [ ] T3 images ≤ 20% of total
- [ ] All 6 categories have 3+ assets
- [ ] Hero is T1 editorial
- [ ] No file over 200KB (JPG) / 300KB (PNG)
- [ ] No duplicate files
- [ ] 30-second scroll test: every scroll window contains ≥1 T1 or strong T2 image

## Reliable Enrichment Sources (by success rate)

1. **App Store / Google Play** — always public, never blocked
2. **Mobbin** — real UI screenshots, unblocked
3. **Vimeo** — campaign videos, embeddable
4. **Behance** — agency case studies (search by brand + agency name)
5. **Brand guidelines pages** — many fintechs publish these publicly

### Sources that frequently block automation
- Brand websites (Cloudflare/CAPTCHA) — flag as "needs manual collection"
- Instagram (login wall) — needs human to save and upload
- Medium (Cloudflare)
- Most editorial sites (cookie consent walls)

For blocked sources: flag as "needs manual collection" in the critique and specify what's needed.

## Versioning
- Never overwrite originals
- Next version = `[brand]-v[N+1]/`
- Original stays live until sign-off on new version
- Git commit message: `"[brand]: visual research rebuild — X images, Y sections"`

## Comms Protocol
- Main thread: protocol decisions, approvals
- Per-brand messages: tag with `[BRAND]` prefix
- Handoff: auditor posts RESEARCH_COMPLETE → builder (Kitt) picks up → QA → Assaf sign-off
- Kittoro does NOT commit to git or deploy — Kitt only

## Benchmark References
- **CashApp** — gold standard for visual richness (71 images, culture + lifestyle sections), score 72
- **Robinhood** — gold standard for brand depth (53 images, personality + layout sections)
- Both pages are read-only reference — do not modify

## Comparison Page
Live at: `https://staging.curiousendeavor.com/etoro-competition/visual-competitive-research/_comparison/`
Source: `/root/.openclaw/workspace/public/etoro-competition/visual-competitive-research/_comparison/index.html`