keyword-cvr-leak

name	keyword-cvr-leak
description	Find Amazon keywords where shoppers search for a benefit your listing doesn't visually demonstrate. Cross-references Reverse ASIN keywords (high search volume) with review-mined drivers (what customers actually buy for) — flagging mismatches as CVR leaks. Use when running `keyword-cvr-leak {ASIN}`, when CVR is below category average, or when planning the listing image stack.

keyword-cvr-leak — Keyword × Visual-Coverage Diagnostic

A specific, high-leverage diagnostic: every high-volume keyword that searches for a benefit, where your listing doesn't visually show that benefit, is bleeding CVR. This skill finds them.

Invocation

keyword-cvr-leak {ASIN}
keyword-cvr-leak {ASIN} --threshold=10000   # min search volume to flag (default 5000)

Output

/tmp/cro-research/{ASIN}-cvr-leaks-{date}.md

Tools Used

n8n MCP wrapper. Plus inputs from prior skills.

Step	Tool	Purpose
Keywords	Keyword_Research_V2_Reverse_ASIN (results_count=200, additional_details=1)	High-volume keywords + benefit signals
Reviews	Get_Product_Reviews 5★ pages 1-3	What customers love about it
Listing	Get_Product_Details (with realtime_data=1)	Current bullets + image set + title

Phase 1 — Pull

Run all 3 in parallel. Use existing asin-deep-research outputs if available.

Phase 2 — Build the Coverage Map

For each keyword above threshold:

1. Extract the benefit/feature implied by the keyword (NLP: "leg elevation pillow" → benefit = leg elevation; "pillow for back pain" → benefit = back pain relief)
2. Check: does the title mention this benefit? (yes/no)
3. Check: does any listing image visually show this benefit? (yes/no — requires inspection of image_urls)
4. Check: does any bullet call it out? (yes/no)

A keyword is a "CVR leak" when:

• Search volume ≥ threshold AND
• Customers in 5★ reviews confirm this benefit matters AND
• Title OR image OR bullet does NOT cover it

Phase 3 — Rank Leaks by Impact

Score each leak: search_volume × (1 if reviews-confirmed else 0.3) × (3 if image-missing, 2 if bullet-missing, 1 if title-missing).

Top 10 = priority fix list.

Phase 4 — Output

# Keyword CVR Leak Analysis — {Title}

**ASIN:** {ASIN} | **Date:** {date} | **Volume threshold:** {N}

## Data Sources

| Source | Records | Status |
|---|---|---|
| Reverse ASIN keywords | 200 | ✅ |
| Reviews (5★) | 30 | ✅ |
| Current listing | 1 ASIN | ✅ |

## Top 10 CVR Leaks (ranked by impact)

| # | Keyword | Vol/mo | Implied Benefit | Reviews-Confirm | Title? | Bullet? | Image? | Recommended Fix |
|---|---------|--------|-----------------|-----------------|--------|---------|--------|------------------|
| 1 | "leg elevation pillow" | 45000 | Leg elevation | ✅ "I use this for elevation" | ❌ | ✅ | ❌ | Add lifestyle slot 3 showing leg-elevation use |
| 2 | ... | ... | ... | ... | ... | ... | ... | ... |

## Coverage Gaps Summary

- **Image gaps** (highest CVR impact): {N} keywords
- **Bullet gaps**: {N} keywords
- **Title gaps**: {N} keywords

## Recommended Next Skills

- Generate the missing visual: `lifestyle-stack-generator {ASIN}` or `infographic-builder {ASIN}`
- Fix the listing copy: `copy-split-test {ASIN}`
- Full CVR fix pipeline: `cvr-leak-fix {ASIN}`

Reference Files

• Vault: CRO-Knowledge-Base/MASTER-CRO-REFERENCE.md §4 ("High-volume keyword for benefit not shown → CVR leak — fix immediately")
• ~/.claude/knowledge/cro-methodology/decision-framework.md

Quality Bar

• At least 10 keywords assessed; top 10 leaks ranked
• Coverage check is honest (verified by inspecting actual title / bullets / images)
• Recommended fix per leak is specific (slot number, image type, or bullet rewrite)
• Reviews-confirmation column filled (otherwise this is just "keyword research")

Common Mistakes

• ❌ Flagging brand keywords as leaks (you're already there)
• ❌ Not checking the actual image content — keyword may be in alt-text
• ❌ Treating low-volume long-tail as leaks (focus on threshold)

Auto-Triggers

• User says "why is CVR low on {ASIN}" / "what keywords am I missing"
• cvr-leak-fix calls it as a sub-step
• /cro-content-plan Phase 2 references it