Amazon Review Intelligence Monitor

Amazon Review Intelligence Monitor

Turn raw Amazon product reviews into complaint themes, evidence quotes, urgency scores, listing actions, ops follow-ups, and seller-response drafts. This Actor analyzes Amazon review records collected by another Apify Actor or supplied inline; it does not scrape Amazon directly.

Instead of returning another spreadsheet of review text, it gives marketplace operators and agencies a clean action dataset: what customers complain about, what they praise, which reviews need attention, and what to fix in the product, listing, support, or fulfillment workflow.

Workflow Hub

See the public review intelligence workflow for the scraper dataset -> analyzer path and links across the review-intelligence Actors. For the first run, use the Amazon review tutorial or the marketplace review action queue use case. The proof GIF shows reviews becoming a marketplace action queue.

What You Learn

• Which reviews are positive, negative, or mixed
• Which themes keep appearing: quality, shipping, fit, setup, price, support, and experience
• Which complaints deserve urgent follow-up
• Which positive reviews can become listing copy, comparison proof, or ad proof
• Which exact quote explains the defect, listing, support, or fulfillment issue
• Whether the next step is listing copy, support, packaging, QA, or monitoring
• What action to take for each review

Use Cases

• Weekly Amazon review monitoring for brands and agencies
• Product-quality, packaging, and fulfillment complaint triage
• Listing-copy improvement from recurring buyer language
• Positive-review proof mining for ads and product detail pages
• Review exports to Google Sheets, Slack, dashboards, or reporting decks

Input

You can provide reviews inline or pass an Apify datasetId from another review scraper.

{
  "productName": "Northstar Standing Desk",
  "sourceName": "Amazon",
  "reviews": [
    {
      "rating": 1,
      "text": "The desktop cracked after two weeks and the replacement process was confusing.",
      "authorName": "Example Customer",
      "date": "2026-05-01"
    }
  ],
  "maxReviews": 100,
  "includeRawReview": true
}

Output

Each dataset item is one analyzed review:

{
  "status": "succeeded",
  "recordIndex": 1,
  "billingEventName": "amazon-review-analyzed",
  "productName": "Northstar Standing Desk",
  "sourceName": "Amazon",
  "rating": 1,
  "sentimentLabel": "negative",
  "sentimentScore": -100,
  "detectedThemes": ["quality", "support"],
  "complaintThemes": ["quality", "support"],
  "urgencyScore": 95,
  "priorityReason": "Operational complaint affects returns, ranking, or buyer confidence.",
  "workflowCategory": "urgent_listing_or_ops_fix",
  "proofQuote": "The desktop cracked after two weeks and the replacement process was confusing.",
  "listingAction": "No listing-copy change until this theme repeats.",
  "operationsAction": "Escalate defect language to product QA or supplier.",
  "recommendedAction": "Escalate seller-support or refund language for Northstar Standing Desk; unresolved support complaints hurt conversion.",
  "analyzedAt": "2026-05-12T12:00:00+00:00"
}

The run also writes a SUMMARY key-value-store record with analyzed counts, sentiment counts, top complaint themes, top urgent reviews, praise quotes, client-report bullets, and the charge event name.

FAQ

Does this scrape Amazon?

No. It analyzes Amazon review records you provide inline or through an Apify dataset from another Actor. It is an intelligence layer, not an Amazon scraper.

What input do I need for the first run?

Use the Store example with one or more review records. At minimum, each record should include review text, and ratings improve sentiment and urgency scoring.

What do I get back?

One dataset item per analyzed review, including sentiment, detected themes, urgency, proof quote, listing action, operations action, and recommended next step. The run also writes a SUMMARY record.

Can I use data from another review scraper?

Yes. Pass a datasetId from another Actor or paste inline records. The analyzer recognizes common review fields such as text, reviewText, rating, asin, sku, variant, and verifiedPurchase.

How much does it cost?

The configured paid event is amazon-review-analyzed at $0.004 per successfully analyzed review. Check the live Store pricing panel before large runs.

Pricing

Default monetization model: pay per event.

Recommended chargeable event:

• Event name: amazon-review-analyzed
• Event meaning: one successfully analyzed review
• Store price: $0.004 per analyzed review
• Pricing status: active from 2026-05-13T19:04:21Z; verified by private and public paid smokes

Successful rows are pushed only after the charge path allows the event. If Actor.charge fails, the Actor fails closed before returning paid output.

Limitations

• This MVP analyzes review records; it does not scrape Amazon directly.
• Review source schemas vary. The Actor recognizes common fields such as text, reviewText, rating, stars, authorName, date, reviewUrl, asin, sku, variant, and verifiedPurchase.
• Theme detection is deterministic and intentionally explainable. It is built for reliable monitoring, not black-box sentiment theater.

Automation And Agent Use

• Run a review scraper first, then pass its dataset ID to this Actor.
• Schedule weekly analysis for each ASIN, product line, or brand account.
• Send negative high-urgency reviews to Slack or a ticketing queue.
• Append sentimentLabel, detectedThemes, and recommendedAction to Google Sheets for client reporting.

Local Development

python3 -m pip install -r requirements.txt
ACTOR_TEST_PAY_PER_EVENT=true apify run --purge --input-file examples/smoke-input.json