amazon-review-qa-complaint-miner

Production-ready Apify Actor that mines Amazon public product feedback into structured review, qna, theme, and warning rows.

Target buyers

• Ecommerce operators validating product-market fit
• Marketplace analysts monitoring complaint trends
• Product managers looking for missing-feature requests
• Agencies building voice-of-customer reporting pipelines

What it does

• Accepts Amazon ASINs or product URLs
• Fetches public product, review, and question pages over normal HTTP requests only
• Detects blocked pages such as captcha, robot check, and HTTP 403 responses
• Emits warning rows instead of crashing when blocking is encountered
• Supports deterministic offline fixture mode through fixtureHtmlPaths
• Extracts review rows, Q&A rows, and derived theme rows
• Categorizes themes with deterministic keyword rules

Public-data limits

• The actor only works against public HTML and does not log in
• Amazon can rate-limit, geo-gate, or challenge requests with captcha/robot pages
• Structure can vary by marketplace and page template, so extraction is best-effort
• When public pages are blocked, the actor outputs warning rows with evidence instead of failing the run

Input

Supported input fields:

• asins: array of ASIN strings
• urls: array of Amazon product URLs
• marketplaceDomain: domain such as www.amazon.com
• maxReviews: maximum number of reviews to emit per product
• maxQuestions: maximum number of Q&A rows to emit per product
• ratingFilter: optional minimum star rating number
• startDate: optional ISO date lower bound
• endDate: optional ISO date upper bound
• fixtureHtmlPaths: optional local HTML paths for offline deterministic mode

Example:

{
  "asins": ["B0TEST1234"],
  "marketplaceDomain": "www.amazon.com",
  "maxReviews": 10,
  "maxQuestions": 10,
  "ratingFilter": 1,
  "startDate": "2024-01-01",
  "endDate": "2024-12-31",
  "fixtureHtmlPaths": [
    "tests/fixtures/product.html",
    "tests/fixtures/reviews.html",
    "tests/fixtures/questions.html"
  ]
}

See examples/sample-input.json.

Output

The actor pushes a flat dataset containing mixed row types:

• review: normalized review data
• qna: normalized product question and answer data
• theme: mined complaint, praised-feature, and missing-feature insights
• warning: non-fatal fetch or block detection event

Each theme row includes:

• themeType
• category
• urgency
• quote
• rating
• date
• sourceUrl
• evidence

See examples/sample-output.json.

Local commands

Install dependencies:

$npm install

Run tests:

$npm test

Run deterministic smoke test:

$npm run smoke

Apify run

The actor entrypoint is src/main.js and writes results with Actor.pushData.