Kunze's Ad Setup Skill

This skill implements Kunze's ecosystem-driven PPC launch strategy from the LegacyX FBA course. It creates a complete 5-campaign Sponsored Products launch package using Amazon Bulksheets 2.0 format, ready for upload to Amazon Ads Console.

User Inputs Required

When the user activates this skill, collect these inputs:

Input	Required	Example	Notes
Main keyword	Yes	bamboo cutting board	The primary search term for the product niche
Product SKU	Yes	BB-CUT-001	The user's seller SKU for the product ad
Budget level	No	conservative or aggressive	Defaults to conservative if not specified
Product name	No	Bamboo Cutting Board	Used in campaign naming. If not given, derive from keyword

Strategy Overview (Kunze's Method)

Kunze's approach is ecosystem-driven using phrase and broad match to cast a wider net and capture how people actually shop on Amazon. Key principles:

• Broad/phrase match focused — captures keyword variations Amazon associates with your product
• 1 ad group per campaign — Amazon is bad at distributing budget across multiple ad groups
• Max 5 keywords per keyword campaign — prevents budget dilution across too many keywords
• Dynamic bids down only — conservative bidding that reduces bids when conversion is unlikely
• Keyword stacking — mix 1 high-volume keyword with a few mid/low-volume keywords per campaign
• Gap analysis — find keywords competitors don't rank well for (higher opportunity, lower CPC)

Workflow

Phase 1: Market Research & Competitor Collection

Run research_products to identify the top competitors and collect ASINs:

research_products(keyword="<main_keyword>", focus="balanced", product_limit=20)

From the results, collect:

1. Top 10-15 ASINs sorted by monthly_revenue descending
2. Market-level data: avg_price, avg_reviews, search_volume, opportunity_score, market_grade
3. Category information for the Category targeting campaign
4. Note the dominant brand and brand concentration

Store these for later phases:

• competitor_asins: Array of 10-15 top ASINs
• market_data: The market-level summary object
• product_list: Full product array with revenue, reviews, price per product

Phase 2: Batched Keyword Research (SUB-AGENT DELEGATION)

CRITICAL: This phase MUST be delegated to sub-agents to protect the main context window.

Keyword research returns very large result sets. Running it in the main context will consume too much of the context window. Instead, split the competitor ASINs into batches of 3-4 and delegate each batch to a sub-agent.

Batching rules:

• Maximum 3-4 ASINs per batch
• Use limit=50 per batch to keep results manageable
• Use min_search_volume=300 to filter out ultra-low-volume keywords
• Run batches in parallel for speed
• Each sub-agent should return a filtered, deduplicated keyword list

Sub-agent prompt template for each batch:

You are a keyword research assistant. Run the following keyword research and return a clean filtered list.

Run this tool call:
amazon_keyword_research(asins=["ASIN1", "ASIN2", "ASIN3"], limit=50, min_search_volume=300)

From the results, extract the top 20 keywords sorted by search_volume descending.

Return results as a STRUCTURED LIST with one keyword per line in this exact format:
KEYWORD | search_volume | cpc | purchase_rate | relevance_score | ASIN1_rank | ASIN2_rank | ASIN3_rank

Example:
bamboo cutting board | 125235 | 1.28 | 5.33 | 892 | 12 | 45 | -
cutting boards for kitchen | 321719 | 2.39 | 3.61 | 756 | 8 | - | 23

Use "-" if an ASIN is not ranking. Use pipe delimiters (|) for easy parsing.

Filter OUT keywords that:
- Have search_volume below 300
- Are brand names of competitors (e.g., Epicurean, Boos, specific brand names)
- Are completely irrelevant to the product category
- Are single generic words (e.g., "board", "wood", "bamboo")

Also report on separate lines:
TOTAL_FOUND: [number]
PASSED_FILTERS: [number]
TOP_PURCHASE_RATE: [keyword1], [keyword2], [keyword3]

Launch sub-agents like this (example with 12 ASINs = 3 batches):

# Launch all 3 in parallel using the Agent tool:
Agent(subagent_type="general-purpose", name="kw-batch-1", prompt="<template with ASINs 1-4>")
Agent(subagent_type="general-purpose", name="kw-batch-2", prompt="<template with ASINs 5-8>")
Agent(subagent_type="general-purpose", name="kw-batch-3", prompt="<template with ASINs 9-12>")

Phase 3: Build Keyword Master List

After all sub-agent batches return, merge and process their results:

3a. Merge & Deduplicate

• Combine all keyword arrays from all batches
• Deduplicate by keyword text (case-insensitive)
• For duplicates: keep the entry with highest search volume, merge competitor rank data from all appearances

3b. Overlap Scoring

• Keywords appearing in multiple batches are stronger signals
• Add an overlap_count field (how many batches contained this keyword)
• Keywords with overlap_count >= 2 get priority

3c. Gap Analysis

• For each keyword, check competitor ranks
• Gap keyword: majority of competitors have rank > 50 or don't rank at all
• Opportunity score = search_volume * (1 / avg_competitor_rank) — higher = better opportunity
• Flag keywords where most competitors rank poorly as "gap opportunities"

3d. Final Filtering & Categorization Categorize keywords into tiers:

• High volume: search_volume >= 5,000
• Mid volume: search_volume 1,000 - 4,999
• Low volume: search_volume 300 - 999

3e. Select Campaign Keywords Apply Kunze's keyword stacking rules:

For Broad campaign (5 keywords max):

• 1-2 high-volume keywords
• 2-3 mid-volume keywords
• Prioritize gap opportunity keywords
• Use the cpc/suggested_bid from research as starting bid

For Phrase campaign (5 keywords max):

• Same 5 keywords as Broad campaign OR different 5 from master list
• Using phrase match with these keywords provides tighter targeting
• Use the cpc/suggested_bid from research as starting bid

For ASIN targeting campaign:

• Default: Top 15 ASINs by monthly revenue from Phase 1
• Expand to 20+ only if the user provides additional specific competitor ASINs
• Bid at or slightly below the average CPC from keyword research

For Category targeting campaign:

• Use the product category identified in Phase 1
• Target the broadest relevant category

Phase 4: Generate Amazon Bulksheets 2.0 CSV Files

IMPORTANT: This phase can also be delegated to a sub-agent if context is getting large.

Generate a single CSV file containing all 5 campaigns. The CSV must use the exact Amazon Bulksheets 2.0 format with these 29 columns in order:

Product,Entity,Operation,Campaign Id,Ad Group Id,Portfolio Id,Ad Id,Keyword Id,Product Targeting Id,Campaign Name,Ad Group Name,Start Date,End Date,Targeting Type,State,Daily Budget,sku,asin,Ad Group Default Bid,Bid,Keyword Text,Match Type,Bidding Strategy,Placement,Percentage,Product Targeting Expression,Audience ID,Shopper Cohort Percentage,Shopper Cohort Type

Campaign naming convention: [Product Name] - [Type]

• Example: Bamboo Cutting Board - Auto, Bamboo Cutting Board - Broad, etc.

Date format: YYYYMMDD (use today's date for Start Date)

Budget and bid settings by level:

Setting	Conservative	Aggressive
Daily Budget per campaign	10	20
Keyword bids	At suggested CPC	10% above suggested CPC
Auto campaign default bid	0.50	0.75
ASIN targeting bid	At avg CPC	At avg CPC
Category targeting bid	At avg CPC	At avg CPC

CRITICAL: Before generating the CSV, create the output directory:

mkdir -p ./campaigns

Campaign 1: Auto Campaign

Structure: Campaign + Ad Group + Product Ad + 4 Auto Targeting rows (7 rows total)

Always include the 4 auto targeting expression rows with individual bids so you can control and optimize each targeting type independently:

• close-match — keywords closely related to your product
• loose-match — keywords loosely related to your product
• substitutes — shoppers viewing similar products
• complements — shoppers viewing complementary products

Each targeting type gets its own Product Targeting row with the expression in the Product Targeting Expression column and an individual bid.

Campaign 2: Manual Broad Match

Structure: Campaign + Ad Group + Product Ad + 5 Keywords (8 rows total)

• Targeting Type: Manual
• 5 keywords with Broad match type
• Individual bids per keyword at their CPC from research

Campaign 3: Manual Phrase Match

Structure: Campaign + Ad Group + Product Ad + 5 Keywords (8 rows total)

Same structure as Broad but with Phrase match type. Can use the same 5 keywords or a different set of 5 from the master list.

Campaign 4: Category Targeting

Structure: Campaign + Ad Group + Product Ad + 1 Product Targeting (4 rows total)

• Targeting Type: Manual
• Product Targeting Expression: category="[CATEGORY_ID]"
• Bid at average CPC

Note on Category ID: The category ID must come from the user or from Amazon's category taxonomy. If the user doesn't provide it, include a placeholder [CATEGORY_ID] and instruct the user to find it by:

1. Go to Amazon Ads Console > Campaign Manager > Create Campaign > Manual Targeting > Product Targeting
2. Browse or search categories to find the relevant one
3. The category ID is the numeric ID shown (e.g., category="5524098011")
4. Alternatively, download an existing campaign's bulksheet that has category targeting — the ID will be in the Product Targeting Expression column

Campaign 5: ASIN Targeting

Structure: Campaign + Ad Group + Product Ad + 15 Product Targeting rows (18 rows total)

• Targeting Type: Manual
• Default: Top 15 ASINs by monthly revenue from Phase 1. Expand to 20+ only if user provides additional specific competitor ASINs.
• Product Targeting Expression: asin="BXXXXXXXXXX" per ASIN
• Bid at average CPC

CSV escaping for ASIN expressions: The value asin="B071NV54JD" contains double quotes. Use Python's csv.writer which handles escaping automatically.

CSV Generation Rules

1. Save as UTF-8 CSV in the project directory: ./campaigns/[keyword]-kunze-campaigns.csv
2. All 5 campaigns in a single file — Amazon supports multiple campaigns per upload
3. No dollar signs or currency symbols — numeric values only for budget and bid fields
4. Empty fields — leave truly empty (no space, no quote), just consecutive commas
5. Campaign ID = Campaign Name — for new campaigns, use the same text-based name in both fields
6. Ad Group ID = Ad Group Name — same rule applies
7. Dates — format YYYYMMDD, use today's date

CSV Validation (REQUIRED)

After generating the CSV, always verify column counts before presenting to the user:

awk -F',' '{print NF}' ./campaigns/[keyword]-kunze-campaigns.csv | sort | uniq -c

Expected output: All rows should show 29. If any row has a different count, fix it before proceeding.

Also verify:

• Total row count matches expected: header + 7 (Auto) + 8 (Broad) + 8 (Phrase) + 4 (Category) + 18 (ASIN) = 46 rows typical
• Broad keywords have Match Type = Broad, Phrase keywords have Match Type = Phrase
• No empty Campaign Name or Ad Group Name fields

REQUIRED: Use Python csv.writer to generate the CSV. Manual comma counting is error-prone and has caused column misalignment bugs. Use the make_row() helper pattern for reliable output:

import csv
from datetime import date

HEADER = ['Product','Entity','Operation','Campaign Id','Ad Group Id','Portfolio Id','Ad Id','Keyword Id','Product Targeting Id','Campaign Name','Ad Group Name','Start Date','End Date','Targeting Type','State','Daily Budget','sku','asin','Ad Group Default Bid','Bid','Keyword Text','Match Type','Bidding Strategy','Placement','Percentage','Product Targeting Expression','Audience ID','Shopper Cohort Percentage','Shopper Cohort Type']
COLS = len(HEADER)  # 29
today = date.today().strftime('%Y%m%d')

def make_row(**kwargs):
    row = [''] * COLS
    field_map = {
        'product': 0, 'entity': 1, 'operation': 2, 'campaign_id': 3,
        'ad_group_id': 4, 'portfolio_id': 5, 'ad_id': 6, 'keyword_id': 7,
        'pt_id': 8, 'campaign_name': 9, 'ad_group_name': 10, 'start_date': 11,
        'end_date': 12, 'targeting_type': 13, 'state': 14, 'daily_budget': 15,
        'sku': 16, 'asin': 17, 'ag_default_bid': 18, 'bid': 19,
        'keyword_text': 20, 'match_type': 21, 'bidding_strategy': 22,
        'placement': 23, 'percentage': 24, 'pt_expression': 25,
        'audience_id': 26, 'shopper_pct': 27, 'shopper_type': 28
    }
    for k, v in kwargs.items():
        if k in field_map:
            row[field_map[k]] = str(v)
    return row

rows = []

# Campaign row example:
rows.append(make_row(
    product='Sponsored Products', entity='Campaign', operation='Create',
    campaign_id=name, campaign_name=name, start_date=today,
    targeting_type='Auto', state='Enabled', daily_budget=budget,
    bidding_strategy='Dynamic bids - down only'
))

# Ad group row example:
rows.append(make_row(
    product='Sponsored Products', entity='Ad group', operation='Create',
    campaign_id=name, ad_group_id=ag_name, ad_group_name=ag_name,
    state='Enabled', ag_default_bid=default_bid
))

# Keyword row example (for Broad/Phrase campaigns):
rows.append(make_row(
    product='Sponsored Products', entity='Keyword', operation='Create',
    campaign_id=name, ad_group_id=ag_name, state='Enabled',
    bid=keyword_bid, keyword_text=keyword, match_type='Broad'
))

# ASIN targeting row example:
rows.append(make_row(
    product='Sponsored Products', entity='Product targeting', operation='Create',
    campaign_id=name, ad_group_id=ag_name, state='Enabled',
    bid=asin_bid, pt_expression=f'asin="{asin}"'
))

with open(output_path, 'w', newline='') as f:
    writer = csv.writer(f)
    writer.writerow(HEADER)
    writer.writerows(rows)

NEVER use inline CSV string concatenation with commas. Always use the make_row() helper above.

Phase 5: Pro-Tip Low-Bid Auto Campaign (Optional)

Per Kunze's pro tip: create an additional auto campaign with very low bids ($0.20-$0.25) to catch high-ROAS traffic during low-competition hours. Only include this if the user selected aggressive budget level or explicitly requests it.

Structure: Campaign + Ad Group + Product Ad (3 rows)

• Campaign Name: [ProductName] - Auto Low Bid
• Targeting Type: Auto
• Daily Budget: $5
• Default bid: $0.22
• Use make_row() helper — same pattern as Campaign 1 but with lower bids

Phase 6: HTML Summary Artifact

After generating the CSV, present a comprehensive HTML summary to the user. This should be displayed inline in the chat as an HTML artifact.

HTML template structure:

<html>
<head>
<style>
  body { font-family: -apple-system, BlinkMacSystemFont, 'Segoe UI', sans-serif; max-width: 900px; margin: 0 auto; padding: 20px; background: #f8f9fa; }
  .header { background: linear-gradient(135deg, #232f3e 0%, #37475a 100%); color: white; padding: 24px; border-radius: 12px; margin-bottom: 24px; }
  .header h1 { margin: 0 0 8px 0; font-size: 24px; }
  .header .subtitle { opacity: 0.85; font-size: 14px; }
  .card { background: white; border-radius: 10px; padding: 20px; margin-bottom: 16px; box-shadow: 0 1px 3px rgba(0,0,0,0.08); }
  .card h2 { margin-top: 0; font-size: 18px; color: #232f3e; border-bottom: 2px solid #ff9900; padding-bottom: 8px; }
  .scorecard { display: grid; grid-template-columns: repeat(auto-fit, minmax(180px, 1fr)); gap: 12px; }
  .metric { background: #f0f2f5; padding: 14px; border-radius: 8px; text-align: center; }
  .metric .value { font-size: 24px; font-weight: 700; color: #232f3e; }
  .metric .label { font-size: 12px; color: #666; margin-top: 4px; }
  table { width: 100%; border-collapse: collapse; font-size: 13px; }
  th { background: #232f3e; color: white; padding: 10px 12px; text-align: left; }
  td { padding: 8px 12px; border-bottom: 1px solid #e8e8e8; }
  tr:hover td { background: #fff8f0; }
  .badge { display: inline-block; padding: 2px 8px; border-radius: 12px; font-size: 11px; font-weight: 600; }
  .badge-high { background: #e8f5e9; color: #2e7d32; }
  .badge-mid { background: #fff3e0; color: #e65100; }
  .badge-low { background: #fce4ec; color: #c62828; }
  .badge-gap { background: #e3f2fd; color: #1565c0; }
  .campaign-row { display: flex; justify-content: space-between; align-items: center; padding: 12px 0; border-bottom: 1px solid #eee; }
  .campaign-name { font-weight: 600; }
  .campaign-details { font-size: 13px; color: #666; }
  .budget-summary { background: #fff8e1; border-left: 4px solid #ff9900; padding: 16px; border-radius: 0 8px 8px 0; margin-top: 16px; }
  .file-link { background: #e8f5e9; padding: 12px 16px; border-radius: 8px; font-family: monospace; font-size: 13px; }
</style>
</head>
<body>
  <div class="header">
    <h1>Kunze's Ad Setup: [Product Name]</h1>
    <div class="subtitle">PPC Launch Package — [Date] — Keyword: "[main_keyword]"</div>
  </div>

  <!-- Market Scorecard -->
  <div class="card">
    <h2>Market Scorecard</h2>
    <div class="scorecard">
      <div class="metric"><div class="value">[search_volume]</div><div class="label">Search Volume</div></div>
      <div class="metric"><div class="value">$[avg_price]</div><div class="label">Avg Price</div></div>
      <div class="metric"><div class="value">[avg_reviews]</div><div class="label">Avg Reviews</div></div>
      <div class="metric"><div class="value">[opportunity_score]/100</div><div class="label">Opportunity Score</div></div>
      <div class="metric"><div class="value">[market_grade]</div><div class="label">Market Grade</div></div>
      <div class="metric"><div class="value">$[avg_cpc]</div><div class="label">Avg CPC</div></div>
    </div>
  </div>

  <!-- Keyword Master List (top 20) -->
  <div class="card">
    <h2>Keyword Master List (Top 20)</h2>
    <table>
      <tr><th>Keyword</th><th>Volume</th><th>CPC</th><th>Tier</th><th>Gap?</th><th>Selected For</th></tr>
      <!-- Repeat for each keyword -->
      <tr>
        <td>[keyword]</td>
        <td>[volume]</td>
        <td>$[cpc]</td>
        <td><span class="badge badge-[tier]">[TIER]</span></td>
        <td>[Yes/No]</td>
        <td>[Broad, Phrase, or —]</td>
      </tr>
    </table>
  </div>

  <!-- Campaign Preview -->
  <div class="card">
    <h2>Campaign Preview (5 Campaigns)</h2>
    <!-- Repeat for each campaign -->
    <div class="campaign-row">
      <div>
        <div class="campaign-name">[Campaign Name]</div>
        <div class="campaign-details">[Type] | [X] entities | Bid: $[bid] | Budget: $[budget]/day</div>
      </div>
    </div>
  </div>

  <!-- Budget Summary -->
  <div class="card">
    <h2>Budget Estimate</h2>
    <div class="budget-summary">
      <strong>[Conservative/Aggressive] Plan</strong><br>
      Daily: $[total_daily] across 5 campaigns<br>
      Monthly estimate: ~$[monthly_estimate]<br>
      Starting bids: [at/above] suggested CPC ($[avg_cpc])
    </div>
  </div>

  <!-- File Output -->
  <div class="card">
    <h2>Generated Files</h2>
    <div class="file-link">campaigns/[keyword]-kunze-campaigns.csv</div>
    <p style="font-size:13px;color:#666;margin-top:8px;">Upload this CSV to Amazon Ads Console > Bulk Operations > Upload</p>
  </div>
</body>
</html>

Sub-Agent Delegation Summary

To manage context window limits, delegate these phases to sub-agents:

Phase	Delegate?	Why	Sub-agent Type
Phase 1: Market Research	No	Small result, needed for orchestration	Main agent
Phase 2: Keyword Research	YES	Large results, 3-4 parallel batches	general-purpose per batch
Phase 3: Keyword Processing	Optional	Only if main context is getting large	general-purpose
Phase 4: CSV Generation	Optional	Only if main context is getting large	general-purpose
Phase 5: Low-Bid Auto	No	Trivial, just extra rows	Main agent
Phase 6: HTML Summary	No	Must be presented in main chat	Main agent

Minimum sub-agents needed: 3-4 (one per keyword research batch) Maximum sub-agents: 6-7 (if also delegating processing and CSV generation)

Bulksheet Format Quick Reference

29 Column Headers (in order)

Product | Entity | Operation | Campaign Id | Ad Group Id | Portfolio Id | Ad Id | Keyword Id | Product Targeting Id | Campaign Name | Ad Group Name | Start Date | End Date | Targeting Type | State | Daily Budget | sku | asin | Ad Group Default Bid | Bid | Keyword Text | Match Type | Bidding Strategy | Placement | Percentage | Product Targeting Expression | Audience ID | Shopper Cohort Percentage | Shopper Cohort Type

Entity Types & Required Fields

Entity	Required Fields (besides Product, Entity, Operation)
Campaign	Campaign Id, Campaign Name, Start Date, Targeting Type, State, Daily Budget, Bidding Strategy
Ad group	Campaign Id, Ad Group Id, Ad Group Name, State, Ad Group Default Bid
Product ad	Campaign Id, Ad Group Id, State, sku (sellers) or asin (vendors)
Keyword	Campaign Id, Ad Group Id, State, Bid, Keyword Text, Match Type
Product targeting	Campaign Id, Ad Group Id, State, Bid, Product Targeting Expression
Negative keyword	Campaign Id, Ad Group Id, State, Keyword Text, Match Type (negativeExact/negativePhrase)
Campaign negative keyword	Campaign Id, State, Keyword Text, Match Type (negativeExact/negativePhrase)
Bidding adjustment	Campaign Id, Placement, Percentage

Key Values

Field	Allowed Values
Product	Sponsored Products
Operation	Create, Update, Archive
Targeting Type	Auto, Manual
State	Enabled, Paused
Match Type (keywords)	Exact, Broad, Phrase
Match Type (negative)	negativeExact, negativePhrase
Bidding Strategy	Dynamic bids - down only, Dynamic bids - up and down, Fixed bid
Placement	placement top, placement product page, placement rest of search(BETA)
Product Targeting (ASIN)	asin="BXXXXXXXXXX"
Product Targeting (category)	category="XXXX"
Product Targeting (auto)	close-match, loose-match, substitutes, complements
Date format	YYYYMMDD
Budget/Bid	Numeric only, no symbols, use decimal point

Tool Reference

Tool	When to Use
research_products(keyword, focus="balanced", product_limit=20)	Phase 1: Market scan and competitor ASIN collection
amazon_keyword_research(asins=[...], limit=50, min_search_volume=300)	Phase 2: Keyword research (run in sub-agents, 3-4 ASINs per batch)
Agent(subagent_type="general-purpose")	Phase 2-4: Delegate keyword research batches and heavy processing

Error Handling

• Keyword research returns too many results: Reduce limit to 30 or increase min_search_volume to 500
• No competitor ASINs found: Ask user for specific competitor ASINs they want to target
• Category ID unknown: Leave as placeholder [CATEGORY_ID] in CSV with instructions for user to fill in from Amazon Ads console
• Sub-agent fails: Retry the failed batch once. If it fails again, skip that batch and note it in the summary
• Less than 5 good keywords found: Use what's available. A campaign can have fewer than 5 keywords — it just can't have more