SuperAPI AI Directory

commerce-marketplace-coupang-ads

Marketing & growth

View repository

Updated 10-15 23:25

About this skill

Coupang full ad set optimization (Marketplace Coupang Ads)

Overview

A dedicated skill that analyzes and optimizes a Coupang seller's ad operations with a single line of natural language input. Wraps into a skill the real-world know-how of a Korean seller who went from 7,000 in monthly sales to 360 million in 6 months (ad spend-to-sales ratio from 30% to 11.2%).

One-line responsibility: Coupang ad report input → automatic classification of 3 campaign types + separation of search/non-search-area sales + automatic end ROAS calculation + per-product decision branching.

The uniqueness of Coupang ads

Area Search area Non-search area
Placement Keyword search results (middle/top/banner) Recommendations on main, product, cart, and order-completion pages
CPM About 10,000 won 약 60원
CPC Expensive Very cheap
Click-through rate·Conversion rate High Significantly low
Impressions Few Enormously many

Key insight: The search area and non-search area have a 167-fold difference in CPM. Even for the same product, you must separately analyze which area it sells well in.

3 campaign types

Campaign Purpose When to apply Key settings
AI Smart Ad Finding products that sell well New sellers·new product launches 1 campaign for all product options + AI automatic distribution, minimum daily budget of 30,000 won
Sales Optimization campaign Separately isolate products that sell well in the non-search area After securing data following the AI Smart Ad Ad operation method: Automated operation-Sales Optimization
Manual Keyword campaign Enter keywords for products that sell well in the search area After checking search keyword conversions following the AI Smart Ad Ad operation method: Direct input

Principle: A campaign = a storage box. 1 campaign per productis the cleanest (the system does not get confused).

3 types of automated rules

Rule 1: Preventing Coupang golden-time budget-shortage ad suspension

  • If more than 70% of the daily budget is used before 6 PM, automatically increase the daily budget by 40%
  • Recommended: Use only when ROAS is inconsistent (budget changes affect AI learning)
  • Pro tip: Budget 10,000 won + 500% daily budget increase → 10,000 won all day, 50,000 won after 6 PM

Rule 2: Increase budget while maintaining ad return to grow sales scale

  • Automatically increase daily budget by 40% when ROAS is 350% or higher
  • Recommended: Only when ROAS is good but raising the budget is ambiguous

Rule 3: Minimum standards for ad operations (alert/pause when ad spend exceeds sales)

  • Alert or pause ads when the 7-day average ROAS is below 100%
  • Recommended: Always keep on (guardrail to prevent ad spend > sales)

The true meaning of target ROAS (= target advertising return)

Target ROAS is a bid adjustment role, not a guarantee of return.

Setting AI behavior Result
Target 1000% (high) Avoids expensive slots, only targets cheap 100-won clicks Low CPC, looks efficient but doesn't get impressions at all
Target 400% (Coupang recommended) Balanced bidding Recommended starting value
Target 200% (low) Boldly bids even on expensive 1,000-won slots Many impressions and clicks but ad spend explodes

Principle: AI cannot change average order value or conversion rate. It can only adjust CPC.

End ROAS (break-even ROAS) calculation

The minimum advertising return at which you do not incur a loss. For the end-ROAS formula, refer to moai-seller:commerce-margin-calculator references/end-roas.md.

Summary: 엔드 ROAS = (판매가 ÷ 마진) × 100. For example (10,000-won product, cost + shipping 8,000 won → margin 2,000 won), end ROAS is 500%.

Coupang ROAS adjustment: Coupang's on-screen ROAS is the amount excluding VAT → in reality it is 10% lower. ROAS shown as 200% = actually 180%. (For adjustment details, see moai-seller:commerce-margin-calculator references/end-roas.md)

Ad efficiency = CPC × conversion rate × average order value

Lever Improvement method
Lower CPC Exclude keywords, adjust target ROAS
Increase conversion rate Product detail page, reviews, price (commerce-detail-page-copy·commerce-jtbd-personalinked to)
Increase average order value Bundle composition, handle products with high average order value

Principle: If you greatly improve even one, ad efficiency changes dramatically. If conversion rate goes from 1% to 2%, sales double.

Workflow — decision branches by product

After analyzing the ad report, apply the following branches to each product:

[Step 1] 엔드 ROAS 계산 (모든 상품)
       ↓
[Step 2] 실제 ROAS와 엔드 ROAS 비교
       ↓
   ┌───────────────────────────────────────────────────────┐
   │ A. ROAS > 엔드 ROAS × 1.5    → 별도 캠페인 + 예산 증액   │
   │                                (예: 4번 상품 ROAS 652%)  │
   │ B. ROAS < 엔드 ROAS          → 살릴 수 있는지 진단      │
   │     ├─ CTR 높음 + CVR 낮음   → 상세페이지·리뷰·가격 개선 │
   │     ├─ CTR 낮음 + CVR 높음   → 썸네일·로켓배송·무료배송 │
   │     └─ 둘 다 낮음            → 광고 제외 (다른 상품 집중) │
   └───────────────────────────────────────────────────────┘
       ↓
[Step 3] 검색영역 vs 비검색영역 매출 분리
       ↓
   ┌───────────────────────────────────────────────────────┐
   │ A. 비검색영역 매출이 더 큼   → 매출최적화 캠페인          │
   │ B. 검색영역 매출이 더 큼     → 수동키워드 캠페인          │
   │                                (잘 팔린 키워드 모아서)    │
   └───────────────────────────────────────────────────────┘
       ↓
[Step 4] 캠페인 분리 시 사전 작업: 리뷰 최소 작업
        - 사진 후기 1개 + 글 후기 1개라도 (전환율에 큰 영향)

Additional settings (Coupang ad full set)

Setting Recommended value Reason
Automatically apply discount coupons ON 0 won out of pocket (Coupang bears 100%). The lower the target ROAS, the higher the chance a coupon is attached
Keyword exclusion Register up to 500 keywords that only eat up ad spend Immediately correct AI learning direction
Audience Plus OFF Exposure on SNS/portals outside Coupang has low efficiency. AI Smart Ads are automatically ON → must turn OFF with the edit button after campaign creation
Daily budget Minimum 30,000 won 30,000 won gets far more impressions than 10,000 won (10,000 won is under 100,000, 30,000 won is 300,000+)
Campaign name 1. AI스마트광고 400% Distinguish from other campaigns + specify target ROAS

Trigger keywords

  • "Analyze Coupang ads", "AI smart ad setup", "Revenue optimization campaign"
  • "Calculate end ROAS", "Tell me break-even ROAS"
  • "Separate Coupang search area and non-search area"
  • "Apply 3 types of Coupang automated rules"
  • "Register keyword exclusions", "Turn off Audience Plus"

Input·Output

Input:

  • Coupang ad report (Excel, revenue growth ad report, daily + campaign>ad group>product>keyword)
  • Or margin calculator result Excel (commerce-margin-calculator output)
  • Product cost·shipping fee·commission information (for end ROAS calculation)

Output:

  1. ROAS·CTR·CVR·end ROAS table by product
  2. Search area/non-search area revenue separation table
  3. Decision by product (separate campaign·diagnosis·exclusion)
  4. Recommended campaign separation plan (revenue optimization·manual keyword)
  5. Recommend applying 3 types of automated rules ON/OFF
  6. Keyword exclusion candidate list

Chaining

[운영 루틴] commerce-morning-brief (전체 브리핑 또는 주문 요약 모드)
[전략·카피] commerce-market-research → commerce-jtbd-persona → commerce-detail-page-copy → commerce-product-naming
[광고 운영] commerce-marketplace-coupang-ads ← ← 본 스킬 ← ←
            ↓ (광고 분석 결과 기반)
        commerce-detail-page-copy --mode diagnose (CTR/CVR 진단)
        cs-channel-message (NCM 15종 재생성)
        commerce-margin-calculator (엔드 ROAS 재산정)

Pair separation

Pair skill Difference
moai-marketer:marketing-performance-report GA4·Meta·Naver·Kakao multi-channel ROAS integration (this skill specializes in Coupang single channel)
commerce-integrated-strategy 1-page revenue improvement strategy (this skill is ad operation decision branching)
commerce-margin-calculator Dedicated to margin·end ROAS calculation (this skill is dedicated to campaign classification·decision)

Reference materials

  • Coupang ad operation practical know-how (based on 6-month operation cases of Korean sellers)
  • commerce-margin-calculator end ROAS formula (moai-seller:commerce-margin-calculator references/end-roas.md)