b2b-customs-data
海关数据分析 — 进出口记录查询、采购商筛选、市场趋势分析
SKILL.mdskills/b2b-customs-data/SKILL.md
GitHub海关数据分析（Customs Data Analysis）
从进出口贸易记录中筛选高价值采购商，输出可直接用于触达的目标客户清单。
Phase 1: Input Preparation
两种输入路径：
- 有数据文件（CSV/Excel 导出）→ 直接分析：逐个文件完整读取，保留原始行序与列名
- 无数据文件 → 输出分析方法论 + 数据获取渠道建议（官方海关统计、商业贸易数据库等），不编造数据
简易方案（无海关数据时专用）
【简易方案，非海关数据】
本节是简易方案，不是海关数据本身。 如果用户 / 客户已有海关数据（CSV / Excel 导出、付费数据库），以海关数据为准。 简易方案不得覆盖海关数据，不得在与海关数据矛盾时优先，不得把简易方案的结果写成海关数据。 简易方案只在没有海关数据时用，作用是初步线索与沟通切入点，不作最终采购判断。
适用条件
- 用户只给了公司名（可加产品 / 市场），没有海关数据文件。
- 用户问的是：这家公司有没有从中国进口过产品、进口了什么产品、哪里进口。
搜索公式（公开渠道，逐条执行，不批量抓取）
- 先用 LinkedIn / 官网确认公司全称、官网、地址，再搜，避免只凭简称认错公司。
- Google 搜索（任选组合，不编造）：
"公司全称" + import data"公司全称" + import records"公司全称" + shipmentssite:importyeti.com "公司全称""公司全称" importer + 产品关键词（有产品关键词时加）
- 只读公开页面可见内容，不绕付费墙，不批量翻页抓取，把查询链接给用户，作人工核对。
提取字段（只提取页面上写明的，不推算）
- 收货人全称、发货人（供应商）、产品描述、原产国（是否中国）、日期、港口、来源 URL。
- 不编造数量、金额、频率、年总量。页面没写的字段，标注“页面未写”，不推测填充。
核对与诚实边界（全部是硬性要求）
- 重名、缩写、收货人 vs 通知方容易混，必须用地址 / 官网 / LinkedIn 对上号再认；对不上号的，标注“未核对”，不得认。
- 覆盖面：主要覆盖美国进口公开提单，其他市场覆盖不全。搜不到不等于没进口，必须明说。
- 证据等级：Google 摘要不算证据，必须点开源页面、标来源 URL 与日期，按 [确切] / [推断] 区分。
- 数据溯源：用户提供 CSV / Excel 时注明文件名 + 列名 + 行号（不要求 URL）；公开网页来源注明 URL + 查询日期（查询日期 = 实际检索当天，不是文章发布日期；不得用文件行号冒充 URL）。
- 回答的第一段必须是数据来源声明（在任何结论、转录、人物介绍之前）：「本报告未使用官方海关数据，仅为公开信息初步判断；若有官方海关数据请提供，以官方数据为准」——即使找到了第三方提单索引也必须写。
- 给客户的开发信、草稿里只能写“公开记录显示……”，不能写“你的海关数据显示……”或把推断写成已确认事实。
- 简易方案结果只能进初步筛选 / 沟通切入点 / P3 培育，不得单独支撑 P1；要打 P1，须用户提供海关数据验证后重新按 Phase 4 评分。
- 公司真伪验证不属于本技能（用 b2b-osint），简易方案也不能代替真伪验证。
- 证据分级：只有真正打开原页看到的内容才标 [确切]；搜索引擎摘要 / 索引片段一律标 [推断] 或 [未核对]。每条 [确切] 必须带来源 URL 与查询日期。
- 人物背景检索：目标是人时，先确认姓名 + 公司 + 职务，再搜“姓名 + 公司 + 履历/experience”“姓名 + 前雇主”“当地语言职务词”；以上查询都执行过，才可写“未找到”。履历段必须写出前雇主（找到就写名称与时间；没找到就明写“前雇主检索未命中”），不得只写到现任公司。
- 来源选择：优先官方 / 公开聚合源；公司级明细在付费墙后就如实写“受限”，不得因单一站点打不开就断言“没有数据”。
- 人名引用：第三方联系人 / 员工必须附来源 URL 或标“未核实”；无来源不得写成对方现员工。
- 实体分层：全球 / 集团数据不得当成单一法人实体数据；按实体分层，每层注明实体名、注册号（如适用）、港口（如适用）、统计期。
- 百分比口径：任何百分比必须写分母（票数 / 金额）、口径（全球 / 美国 / 当地）、统计期、来源。
- 品类判断：必须依据 HS 编码 + 产品描述，不得只凭名称推断。
- 输出格式：每条关键证据逐条标 [确切] / [推断] / [未核对]；开头一律声明数据来源等级（未使用官方海关数据时写明“仅为公开信息初步判断，若有官方数据以官方为准”）；结尾必须给优先级归类（简易方案默认 P3；降级或升级都必须写明依据；有官方海关数据验证后按 Phase 4 重评）；不得把部分一致写成“全部对得上”，绝对否定须限定“在可查到的公开渠道里”；不同数据库 / 口径 / 统计期的数字必须分开注明，不得拼成一张表。
Phase 2: Filtering — 四层筛选
Filter 1: Product Match
Include:
- Buyers importing products that match (or complement) user's product line
- Downstream manufacturers using user's product as input
Exclude:
- Products outside user's business scope
- Raw materials that are inputs to user's product (not final product)
Filter 2: Geographic Targeting
Include based on user's target markets:
- Specific countries/regions (North America, Europe, Southeast Asia)
- Avoid: countries with trade restrictions or high tariffs
Exclude:
- Markets user is not targeting
- Regions with regulatory barriers
Filter 3: Volume & Frequency
Include:
- Buyers with regular import patterns (multiple shipments per year)
- Volume sufficient for user's MOQ
Exclude:
- One-time buyers (no repeat business potential)
- Volume below user's MOQ
Filter 4: Company Type
Include:
- Manufacturers (final product buyers)
- Brand owners
- Large distributors
Exclude:
- Small retailers (below threshold)
- Trading companies acting as intermediaries
Phase 3: Buyer Pattern Analysis
For each qualified buyer, analyze:
Purchase Patterns
| Metric | What It Tells You |
|---|---|
| Frequency | How often they buy (monthly/quarterly/annually) |
| Volume trend | Growing, stable, or declining purchases |
| Seasonality | Peak buying seasons |
| Supplier concentration | Do they rely on few or many suppliers? |
| Price sensitivity | Volume vs. price correlations |
Example Analysis (Generic Template)
Company: [Buyer Name]
Country: [Country]
Products imported: [Product categories]
Annual volume: [Estimated value]
Frequency: [X] shipments/year
Typical order size: [Range]
Suppliers: [Number of suppliers] (mostly from [countries])
Patterns:
- Peak season: [Q1/Q2/Q3/Q4]
- Order cycle: [Monthly/Quarterly]
- Last shipment: [Date]
Potential approach:
- Angle: [Based on their supplier concentration/price trends]
- Timing: [Best time to reach out]
- Product focus: [Which of your products fits their pattern]
Phase 4: Priority Scoring
Score each prospect based on:
Scoring Matrix
| Criteria | Weight | Score (1-5) |
|---|---|---|
| Industry match | 25% | How well product aligns |
| Volume potential | 25% | Order size and frequency |
| Geographic fit | 20% | Your ability to serve |
| Accessibility | 15% | Ease of outreach (LinkedIn, email, etc.) |
| Growth trend | 15% | Purchase volume trend |
Priority Classification
| Total Score | Priority | Action |
|---|---|---|
| 4.0 - 5.0 | P1 — Hot | Immediate outreach within 24h |
| 3.0 - 3.9 | P2 — Warm | Personalized outreach within 1 week |
| 2.0 - 2.9 | P3 — Medium | Add to nurture sequence |
| < 2.0 | P4 — Low | Periodic check-ins only |
Phase 5: Output — Target Customer List
Output Format
# B2B Trade Data Analysis Report
## Summary
- Total records analyzed: [X]
- Qualified prospects: [X]
- By priority: P1=[X], P2=[X], P3=[X], P4=[X]
- Geographic distribution: [Chart/Table]
## P1 Prospects (Immediate Action)
### 1. [Company Name]
| Field | Details |
|-------|---------|
| Country | [Country] |
| Products | [Product categories] |
| Est. Annual Volume | [Value] |
| Frequency | [X]x/year |
| Last Purchase | [Date] |
| Key Suppliers | [Countries] |
| Approach Angle | [How to position] |
| Recommended Action | [Specific next step] |
### 2. [Company Name]
[Same structure]
## P2 Prospects (This Week)
[Same structure]
## P3-P4 Prospects (Nurture)
[Condensed list format]
## Market Insights
1. [Key finding about the market]
2. [Key finding about competitor suppliers]
3. [Opportunity identified]
## Appendix: Full Data Table
| Company | Country | Product | Volume | Frequency | Score | Priority |
|---------|---------|---------|--------|-----------|-------|----------|
| [Name] | [Country] | [Product] | [Value] | [Freq] | [X.X] | P1 |
Phase 6: Integration with Outreach
From Data to Action
For P1 prospects, generate:
- Customer Brief: 1-page summary of the prospect
- Customized Outreach: Cold email referencing their specific purchase patterns
- Talking Points: Based on their supplier concentration, price trends, seasonality
Outreach Angle Examples
If they buy from multiple suppliers:
"We noticed you work with several [product] suppliers in [country].
We're a specialized manufacturer focusing on [specific product segment].
Would you be open to exploring if we can offer better [specific advantage]?"
If they have seasonal patterns:
"Public records show peak season in [Q2].
We're reaching out now because we'd like to discuss how we can
support your [Q2] requirements with our [product] capabilities."
If they recently expanded volume:
"Congratulations on your growth in [product category]!
We've helped similar companies scale their [specific need].
Would you be open to a brief call to explore if we're a fit?"
Quality Standards
- Data accuracy: Cross-check key data points (company names, volumes) against multiple rows
- HS code validation: Ensure HS codes are correctly interpreted for product mapping
- Currency consistency: Note currency in value fields; flag inconsistencies
- No assumptions: If a field is ambiguous, note it rather than guess
- Source citation: Always cite the source file and row numbers for key findings
- Column mapping disclosure: 分析报告中首次引用数据时，注明对应的原始文件列名。 例如："进口量数据来自文件中「Total Import QTY」列"。这样用户可以快速判断列解读是否正确。
- Completeness: Include all relevant fields in output, even if values are missing
Common Pitfalls
- Over-relying on volume: Big buyers may already have established suppliers — look for disruption opportunities
- Ignoring frequency: One-time large orders may not indicate ongoing business potential
- Missing seasonality: Outreach timed wrong can kill opportunity before it starts
- Generic outreach: Always customize message based on the specific buyer's patterns
- Not verifying data: Company names may have typos or different spellings — verify before outreach
- Privacy concerns: Customs data may have usage restrictions — ensure compliance