review-analysis

In-Depth Analysis of Amazon Product Reviews

Quick Reference

Step	Action	Description
1. Get API Key	Read `.mcp.json`	Get the Sorftime API key
2. Create Report Directory	mkdir + data subdirectory	Create `review-analysis-reports/{ASIN}_{站点}_{日期}/data/`
3. Get Product Data	`product_detail` API	Verify the ASIN and get product information, save the raw SSE data
4. Get Review Data	`product_reviews` API	Get all review data, save the raw SSE data
5. Parse and Classify Negative Reviews	In-memory processing	Extract 1-3 star reviews, classify by pain point
6. Save Analysis Data	JSON output	Save the negative review analysis data to `data/negative_reviews_analysis.json`
7. Generate Analysis Report	Markdown output	Save the final report to `report.md`

Report Output Structure

review-analysis-reports/
└── {ASIN}_{站点}_{YYYYMMDD}/
    ├── report.md                              # 完整分析报告（Markdown）
    └── data/                                  # 原始数据和分析结果
        ├── raw_product_sse.txt                # 原始产品详情SSE响应
        ├── raw_reviews_sse.txt                # 原始评论SSE响应
        └── negative_reviews_analysis.json     # 差评分析结构化数据

Trigger Conditions

When the user uses /review-analysis command and provides an Amazon product ASIN, start this analysis process.

Invocation Format:

/review-analysis {ASIN} {站点}

Example: /review-analysis B0D9ZTW7PS US

Role Setting

You are aSenior Amazon Product Development ConsultantandCustomer Experience Expertwith 10 years of experience, specializing in mining product pain points and improvement opportunities through user reviews.

Your core task is, based on the providednegative review text, to deeply analyze the product's core pain points and providedirectly actionablesolutions.

Refer to the analysis framework in the documentation, but adjust flexibly according to the actual review content.

Analysis Process (Optimized Version v7.0 - 6-Dimension Analysis Framework)

Step 1: Read the API Key

# 使用 Read 工具读取配置文件
Read("D:/amazon-mcp/.mcp.json")

# 从 JSON 中提取 API 密钥
# 格式: "url": "https://mcp.sorftime.com?key={API_KEY}"

Step 2: Create the Report Directory Structure

# 创建报告目录和数据子目录
REPORT_DIR="D:/amazon-mcp/reports/review-analysis/{ASIN}_{站点}_20260315"
mkdir -p "$REPORT_DIR/data"

Step 3: Get Product Data and Save the Raw Response

Use the Bash tool to call the Sorftime API and save the raw response:

# 获取产品详情并保存原始SSE响应
API_KEY="从.mcp.json中获取的密钥"
curl -s -X POST "https://mcp.sorftime.com?key=${API_KEY}" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"product_detail","arguments":{"amzSite":"US","asin":"ASIN"}}}' \
  > "$REPORT_DIR/data/raw_product_sse.txt"

If it returns "Authentication required" or "Authorization failed"：

Inform the user that the API key is invalid or has expired
Guide the user to visit https://sorftime.com/zh-cn/mcp to get a new key
Update the .mcp.json file

If it returns "No corresponding product found"：

Verify the ASIN format (should be 10 alphanumeric characters)
Try using another marketplace
Prompt the user to confirm whether the product is sold on that marketplace

Step 4: Get Review Data and Save the Raw Response

Use the reviewType: "Negative" parameter to specifically get negative reviews：

# 获取差评（1-3星），保存原始SSE响应
curl -s -X POST "https://mcp.sorftime.com?key=${API_KEY}" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"product_reviews","arguments":{"amzSite":"US","asin":"ASIN","reviewType":"Negative"}}}' \
  > "$REPORT_DIR/data/raw_reviews_sse.txt"

Important Notes：

Sorftime returns data in SSE (Server-Sent Events) format
Data format: event: message\ndata: {JSON}\n\n
If the data exceeds 25KB, it will be automatically saved to a temporary file
Use the Read tool to read the temporary file and get the complete data
reviewType: "Negative" returns only 1-3 star reviews, up to 100 items
reviewType: "Both" returns all reviews, but negative reviews may be diluted

Handling large file responses：

# 如果响应被保存到临时文件，复制到报告目录
cp /path/to/temp/file.txt "$REPORT_DIR/data/raw_reviews_sse.txt"

Step 5: Parse review data and generate analysis JSON

Process review data in memory while generating structured analysis data：

import json

# SSE 数据解析步骤：
# 1. 从响应中提取 "data: " 后的 JSON
# 2. 解析 JSON 获取 result.content[0].text
# 3. text 中包含 Unicode 转义的中文和评论数组
# 4. 从评论数组中过滤 1-3 星评论

# 解析示例：
start_idx = content.find('data: ') + 6
json_str = content[start_idx:]
data = json.loads(json_str)

# 提取评论文本
text = data['result']['content'][0]['text']

# 查找评论数组起始位置
reviews_start = text.find('[{')
reviews_json = text[reviews_start:]
reviews = json.loads(reviews_json)

# 过滤 1-3 星评论
negative_reviews = [r for r in reviews if float(r.get('评星', 5)) <= 3.0]

# 按6大类别归类差评（v7.0 增加服务维度）
pain_points = {
    "电子模块故障": [],
    "结构/组装问题": [],
    "设计/功能缺陷": [],
    "外观/材质问题": [],
    "描述不符": [],
    "服务/物流问题": []
}

Service dimension classification keywords：

# 服务/物流问题 - 优先检查
service_keywords = {
    '收到二手/瑕疵品': ['used', 'gross', 'dirty', 'scratch', 'ear wax', 'dirt', 'opened', 'previous owner'],
    '配件缺失': ['missing', 'no cord', 'no cable', 'no charger', 'no ear tip', 'no accessory'],
    '退换货困难': ['return', 'refund', 'exchange', 'difficult', 'challenging'],
    '客服问题': ['customer service', 'seller', 'vendor', 'support'],
    '物流问题': ['shipping', 'delivery', 'package', 'packaging'],
    '发错货': ['wrong item', 'wrong color', 'wrong size', 'sent wrong']
}

Save analysis data to JSON：

# 使用 Write 工具生成分析数据文件
# 文件路径: $REPORT_DIR/data/negative_reviews_analysis.json

The JSON file structure should include:

Basic product information (title, brand, price, rating)
Pain point category statistics (category, count, percentage, severity)
Detailed analysis of each pain point (root cause, improvement suggestions, customer quotes)
Safety warnings (if any)
Quality metric estimates

Step 6: Generate analysis report

import json

# SSE 数据解析步骤：
# 1. 从响应中提取 "data: " 后的 JSON
# 2. 解析 JSON 获取 result.content[0].text
# 3. text 中包含 Unicode 转义的中文和评论数组
# 4. 从评论数组中过滤 1-3 星评论

# 解析示例：
start_idx = content.find('data: ') + 6
json_str = content[start_idx:]
data = json.loads(json_str)

# 提取评论文本
text = data['result']['content'][0]['text']

# 查找评论数组起始位置
reviews_start = text.find('[{')
reviews_json = text[reviews_start:]
reviews = json.loads(reviews_json)

# 过滤 1-3 星评论
negative_reviews = [r for r in reviews if float(r.get('评星', 5)) <= 3.0]

Step 7: Generate final analysis report

Use the Write tool to generate a complete Markdown report：

报告路径: $REPORT_DIR/report.md

Analysis framework

Pain point classification (6 major categories)

Category	Judgment criteria
1. Structural/assembly issues	Broken parts, seal failure, connector breakage, mounting hole misalignment, difficult assembly, unstable structure
2. Electronic module failure	USB/charging failure, LED not lighting up, APP connection failure, Bluetooth disconnection, function failure, circuit issues
3. Design/functional defects	Unreasonable size, missing functions, complex operation, tactile mismatch, ergonomic issues, inconvenient to use
4. Appearance/material issues	Odor, material allergy, color difference, scratches, rust, poor surface treatment, cheap-looking material
5. Description mismatch	Size expectation deviation, function not matching description, color difference, style inconsistent with pictures, Bluetooth version mismatch
6. Service/logistics issues	Slow customer service response, difficult returns/exchanges, logistics delays, wrong item shipped, missing accessories,received used product/defective product, damaged packaging

Service dimension breakdown

Service dimension issues need further detailed statistics:

Subcategory	Judgment criteria	Severity
Received used/defective product	Reviews mention used, dirty, ear wax, scratch, opened, previous owner	High
Missing accessories	Missing charging cable, ear tips, manual, warranty card, etc.	High
Difficult returns/exchanges	Complicated return process, slow refunds, seller passing the buck, buyer bearing high shipping costs	Medium
Slow customer service response/poor attitude	Customer service not replying, slow replies, bad attitude, unable to solve problems	Medium
Logistics delays/poor packaging	Slow shipping, logistics stagnation, damaged packaging, poor courier service	Low
Wrong item shipped	Wrong color/size/style shipped	Medium

Severity assessment

Level	Judgment criteria
High	Affects core functions or poses safety hazards (such as damage, leakage, allergy, electric shock)
Medium	Affects user experience (e.g., complicated operation, poor feel, size deviation)
Low	Appearance detail issues (e.g., minor scratches, packaging defects, personal preference)

Dual-track solution

For each pain point, provide:

Product/supply chain improvement plan
- Must be specific and executable (e.g., increase seal width from 3mm to 6mm)
- Avoid vague descriptions (e.g., "improve quality" is unacceptable)
Customer service script/Listing optimization suggestions
- Customer service email template (comply with Amazon compliance requirements)
- Listing copy/image improvement suggestions

Report template

# {产品标题} - 评论深度分析报告

> ASIN: {ASIN} | 站点: {站点} | 分析时间: {时间}

---

## 产品基础信息

| 项目 | 内容 |
|------|------|
| 产品标题 | {标题} |
| 品牌 | {品牌} |
| 价格 | ${价格} |
| 评分 | {评分}/5.0 |
| 评论总数 | {总数} |
| 分析样本 | {差评数量} 条 (1-3星) |

---

## 痛点分析汇总

基于 {差评数量} 条差评的深度分析：

### 痛点分布概览

| 排名 | 痛点类别 | 数量 | 占比 | 严重程度 |
|------|----------|------|------|----------|
| 1 | {类别} | {数量} | {占比}% | {高/中/低} |
| 2 | {类别} | {数量} | {占比}% | {高/中/低} |
| ... | ... | ... | ... | ... |

---

## 核心痛点深度分析

### 痛点 #1: {痛点名称}

**类别**: {类别} | **严重程度**: {程度} | **影响**: {数量}条评论 ({占比}%)

#### 客户反馈摘要
> "{典型差评引用1}"
>
> "{典型差评引用2}"

#### 根源分析
- **设计问题**: {分析}
- **生产问题**: {分析}
- **包装问题**: {分析} (如适用)

#### 产品改进建议
1. {具体可执行的改进1}
2. {具体可执行的改进2}

#### 客服回复模板

**Subject**: {邮件主题}

**Dear [Customer Name],**

{完整的邮件内容}

**Best regards,**

[Your Name]
[Brand Name] Customer Success Team

---

[重复其他痛点...]

---

## 给您的产品开发专家建议

### 产品质量改进
- {建议1}
- {建议2}

### 供应链端的"防呆"设计
- {建议1}
- {建议2}

### Listing与营销层面的"预期管理"
- {建议1}
- {建议2}

### 服务与运营优化（如存在服务维度问题）
- **客服培训**: 建立标准话术库，确保24小时内响应差评
- **退换货流程**: 简化退货流程，提供预付运费标签
- **发货质检**: 100%出库质检，杜绝二手/瑕疵品流出
- **配件管理**: 建立配件清单核对机制，确保包装完整
- **物流合作**: 评估物流服务商，选择可靠的配送渠道

---

## 亚马逊差评回复邮件模板库

### 模板类型（根据痛点类别提供）

1. **产品质量问题**（电子模块故障、结构问题、设计缺陷）
2. **服务问题**（收到二手/瑕疵品、配件缺失、退换货困难）
3. **物流问题**（延迟、包装破损、发错货）
4. **描述不符**（功能预期偏差、尺寸颜色差异）

[根据具体产品类型和痛点提供3-5个针对性模板]

### 服务问题专项模板示例

**收到二手/瑕疵品**:

Subject: We sincerely apologize - We will immediately replace it with a brand new product Dear [Customer Name], We are very sorry that you received a defective product. This absolutely does not meet our quality standards. Please contact [support email] immediately, and we will replace it with a brand new product for free, with no need to return the original product. We apologize again! [Brand] Customer Service


---

## 操作建议（避坑指南）

1. **话术避讳**: 严禁使用 "Change your review" 或 "Remove your review"
2. **回复渠道**: 使用亚马逊后台 "Contact Buyer" 功能
3. **时效性**: 1星评价4小时内响应，2星12小时内，3星24小时内
4. **跟进策略**: 首封邮件聚焦解决问题，不主动提补偿

---

*报告生成时间: {时间戳}*
*数据来源: Sorftime MCP*
*分析方法: LLM 整体评论分析*

Amazon compliance requirements

When generating email templates, you must comply with:

Directly requesting deletion/modification of reviews is strictly prohibited
Reviews must not be exchanged for benefits
Use the official Contact Buyer channel
Respond to negative reviews within 24 hours

Supported marketplaces

US, GB, DE, FR, IN, CA, JP, ES, IT, MX, AE, AU, BR, SA

Troubleshooting

API authorization failure

Symptoms: Returns "Authentication required" or "authorization failed"

Solution:

Visit https://sorftime.com/zh-cn/mcp to obtain a new key
Update .mcp.json API key in
Re-run the analysis

Product not found

Symptoms: Returns "corresponding product not found"

Solution:

Check the ASIN format (10 alphanumeric characters)
Confirm whether the marketplace is correct
Try using another marketplace

Chinese garbled text

Symptoms: The returned data contains \u4ea7\u54c1 and other Unicode escapes

Solution:

Python: json.loads() will automatically decode
If there is Mojibake: text.encode('latin-1').decode('utf-8')

Data truncated because it is too large

Symptoms: Returns "Output too large... saved to: {temp_file}"

Solution:

Read the complete data from the temporary file path shown in the prompt
Use the Read tool's offset/limit parameters to read in chunks
Or use the Grep tool to extract specific patterns

Service issue identification

Symptoms: Service-related negative reviews appear frequently in reviews

Service dimension warning thresholds:

Issue type	Warning threshold	Danger threshold
Receiving used/defective items	>2%	>5%
Missing accessories	>1%	>3%
Complaints about difficulty with returns/exchanges	>5%	>10%
Negative customer service reviews	>3%	>7%

Improvement suggestions:

Used/defective item issue: Immediately review the warehouse quality inspection process and consider a product recall
Missing accessories: Check the packaging line and add accessory barcode verification
Difficulty with returns/exchanges: Simplify the return process and provide prepaid shipping labels
Customer service issues: Increase customer service training and establish a 24-hour response mechanism

The number of negative reviews is very small

Symptoms: The product shows several hundred reviews, but only a few negative reviews are returned

Possible causes:

Good product quality: A low negative review rate is a good thing, indicating high customer satisfaction
API limitations: The Sorftime API returns at most 100 reviews
Use reviewType: "Negative": Only 1-3 star reviews are retrieved, so the number will naturally be small

Data analysis recommendations:

If there are fewer than 5 negative reviews: the analysis results are for reference only, and it is recommended to combine them with other data sources
If there are fewer than 10 negative reviews: clearly state the sample size limitation in the report
If there are more than 20 negative reviews: the analysis results have high statistical significance

Supplementary data plan:

Manually check the negative reviews on the Amazon product page
Use other review scraping tools to obtain more data
Combine with customer service records to understand common issues

Best practices

Path handling: In a Windows environment, using forward slashes / or backslashes \ is acceptable, but be consistent
Data saving: All intermediate data must be saved to the data/ subdirectory to ensure traceability and reuse
JSON structure: Analysis data should use a structured JSON format to facilitate subsequent programmatic processing
Error handling: Check the returned results after each step to identify problems in a timely manner
User feedback: When problems occur, clearly inform the user of the cause and solution

Description of intermediate data files

File name	Purpose	Format
`raw_product_sse.txt`	Raw API response for product details	SSE format
`raw_reviews_sse.txt`	Raw API response for review data	SSE format
`negative_reviews_analysis.json`	Structured data for negative review analysis	JSON

Data reuse scenarios

Trend analysis: Compare changes in negative reviews for the same product across different time periods
Competitor comparison: Batch analyze negative review data for multiple products
Quality traceability: Verify the accuracy of analysis conclusions based on raw data
Report generation: Automatically generate Excel/charts based on JSON data

Version of this skill document: v7.0 (6-dimensional analysis framework) | Last updated: 2026-03-15

About this skill

Quick Reference

Report Output Structure

Trigger Conditions

Role Setting

Analysis Process (Optimized Version v7.0 - 6-Dimension Analysis Framework)

Step 1: Read the API Key

Step 2: Create the Report Directory Structure

Step 3: Get Product Data and Save the Raw Response

Step 4: Get Review Data and Save the Raw Response

Step 5: Parse review data and generate analysis JSON

Step 6: Generate analysis report

Step 7: Generate final analysis report

Analysis framework

Pain point classification (6 major categories)

Service dimension breakdown

Severity assessment

Dual-track solution

Report template

Amazon compliance requirements

Supported marketplaces

Troubleshooting

API authorization failure

Product not found

Chinese garbled text

Data truncated because it is too large

Service issue identification

The number of negative reviews is very small

Best practices

Description of intermediate data files

Data reuse scenarios