用 Claude Code 改善电商商品页: 退货原因、尺码表、FAQ 和退货政策检查

电商店铺退货增加时，商品页常常是原因之一。尺码、颜色、材质或退货政策在购买前没有说清。本文说明如何用 Claude Code 把退货原因、评论、问题、尺码表和退货政策整理成发布前检查表。

Primary references used for the Japanese article: Consumer Affairs Agency mail order, return special terms guideline, National Consumer Affairs Center mail-order warning, NRF 2025 returns landscape, Google Merchant Center product data, Google MerchantReturnPolicy structured data, and Shopify returns help. Related internal reading: BigQuery return analysis for retail EC.

要点

• Return reasons are page-edit inputs, not only support logs.
• Claude Code reviews returns, reviews, questions, size charts, policy pages, and merchant data.
• Humans keep price, inventory, return rules, legal wording, health claims, personal data, and final publication.
• Track return rate, pre-purchase questions, product-page CVR, merchant warnings, and return handling time.
• Start with one product and one review table.

商品页容易失败的地方

The common failure is treating product copy as a launch task only. Support later receives fit, color, material, shipping, and return-policy questions, but the old product page stays unchanged.

Another failure is hiding return information in a shared policy page. Shared pages help, but shoppers also need visible signals near the product page and checkout confirmation.

流程: 从退货原因到商品页修改

Collect product URL, return reason CSV, low-rating reviews, pre-purchase questions, size guide, return policy URL, and Merchant Center or Search Console warnings before writing copy.

Input	Check	Page output
Return reasons	size, color, material, shipping, policy	FAQ and notes
Reviews	expectation gap	photos and comparisons
Questions	pre-purchase doubts	Q&A and links
Size guide	measurements and model info	size help
Return policy	conditions, period, fees	visible policy cues
Product data	title, description, price, availability	merchant data fixes

Claude Code 负责什么，人负责什么

Claude Code classifies reasons, drafts FAQ, finds size-chart gaps, compares policy locations, and makes merchant-data review rows.

Humans decide price, inventory, warranty, refund rules, legal wording, advertising claims, personal-data handling, and final publication. Reviews should be redacted before they enter the prompt.

3 个 Use case

Use case 1: Turn return reasons into FAQ

• Input: return reason CSV, product URL, return policy URL, common questions.
• Output: FAQ, caution text, policy link, and suggested placement.
• Human review: return availability, period, shipping fee, refund timing, and personal data.

Use case 2: Fix size guide and photo captions

• Input: size chart, model photos, return reasons, low-rating reviews.
• Output: missing measurements, captions, comparison table, and top-of-page notes.
• Human review: measurements, material, color, claims, and photo accuracy.

Use case 3: Check return policy and merchant data drift

• Input: product page, shared return page, checkout screen, product data CSV, merchant warnings.
• Output: visibility checklist, missing fields, structured-data review rows.
• Human review: legal display, return special terms, advertising claims, and final dashboard values.

可复制的提示词

EC店舗の商品ページ編集担当として作業してください。目的は、返品理由と問い合わせから、購入前に読者が確認できる情報へ戻すことです。

入力:
- 商品URL:
- 商品名:
- 価格:
- 在庫:
- サイズ表URL:
- 返品ポリシーURL:
- 直近30日の返品理由CSV:
- 低評価レビュー:
- 購入前問い合わせ:
- 類似商品:
- Merchant CenterまたはSearch Consoleの警告:

出力:
1. 返品理由の分類表
2. 商品ページへ追記する説明
3. サイズ表の不足
4. FAQ 5問
5. 返品特約・返品ポリシーの表示チェック
6. Merchant Center用の商品データ確認表
7. 公開前に人が見る項目

制約:
- 価格、在庫、返品可否、返金条件を推測しない
- レビュー本文の個人情報を本文へ入れない
- 医療、美容、健康効果を断定しない
- 返品特約は商品ページ、共通ページ、最終確認画面の3か所で確認する

检查代码

This Node.js sample checks one product-page object. It does not decide law, price, or policy. It only returns missing rows before publication.

const productPage = {
  productName: "撥水ライトジャケット",
  returnReasons: ["サイズが大きい", "写真より色が暗い", "返品条件がわからない"],
  sizeGuide: ["着丈", "身幅", "肩幅"],
  faq: ["返品できますか", "雨の日に使えますか"],
  returnPolicyLocations: ["共通ページ", "商品ページ"],
  merchantData: ["title", "description", "price", "availability"],
  humanChecks: ["価格", "在庫", "返品条件", "レビュー内の個人情報"],
};

const requiredReturnReasonGroups = ["サイズ", "色味", "素材", "配送", "返品条件"];
const requiredSizeGuide = ["着丈", "身幅", "肩幅", "袖丈", "モデル身長"];
const requiredPolicyLocations = ["商品ページ", "共通ページ", "最終確認画面"];
const requiredMerchantData = ["title", "description", "price", "availability", "return_policy"];
const requiredHumanChecks = ["価格", "在庫", "返品条件", "レビュー内の個人情報", "景品表示"];

const warnings = [
  ...requiredReturnReasonGroups
    .filter((item) => !productPage.returnReasons.join(" ").includes(item))
    .map((item) => "返品理由の分類が不足: " + item),
  ...requiredSizeGuide
    .filter((item) => !productPage.sizeGuide.includes(item))
    .map((item) => "サイズ表が不足: " + item),
  ...requiredPolicyLocations
    .filter((item) => !productPage.returnPolicyLocations.includes(item))
    .map((item) => "返品特約の表示場所が不足: " + item),
  ...requiredMerchantData
    .filter((item) => !productPage.merchantData.includes(item))
    .map((item) => "Merchant Center向けデータが不足: " + item),
  ...requiredHumanChecks
    .filter((item) => !productPage.humanChecks.includes(item))
    .map((item) => "人の確認項目が不足: " + item),
];

console.log({
  canPublish: warnings.length === 0,
  warnings,
});

Pitfall: 常见错误

Failure	Cause	Fix
Copy gets longer only	return reasons were ignored	classify reasons first
Size chart is unread	no measurement help	add measurement method and model info
Policy is hard to find	shared page only	check product, shared page, and checkout
Review leaks details	raw review pasted	redact and classify
Merchant warnings remain	page and data differ	compare title, description, price, availability, return policy

ROI 信号

Do not watch return rate alone. Put return rate, product-page CVR, pre-purchase questions, return handling time, non-resellable returns, and merchant warnings in one table.

常见问题

Q. Should a high-return product be stopped? A. Not immediately. Split page-fix returns from product-quality returns first.

Q. Is a shared return policy page enough? A. It helps, but product page and checkout visibility still need review.

Q. Can raw reviews go into Claude Code? A. No. Redact names, order numbers, addresses, photos, and individual context.

何时寻求帮助

For team rollout, bring return CSV, review categories, product URLs, size charts, policy pages, and merchant warnings to 培训与咨询. Start with one product and one return-reason table.

实际验证结果

I checked slug, frontmatter, external references, internal link, CTA, code block, and final section. The sample code reports missing material and shipping return groups, sleeve length, model height, checkout policy location, return_policy field, and advertising review. The first action is to classify one product’s returns into five groups and move the missing answers back to the product page.