Review intelligence that drives decisions, not just data dumps.
Dual-source scraping covers up to 100% of written reviews across 14 Amazon marketplaces. AI semantic clustering surfaces what buyers love, what they hate, and what to fix next. Two ready-to-act documents in minutes — not days.
More & Fresher Data.
Dual-source scraping (Primary + Secondary API) hits 85–100% review coverage — far beyond what Helium 10 or Jungle Scout pull. Fresher data, captured the moment new reviews drop.
Smart Analysis.
AI semantic clustering — not keyword search — understands that "handle wobbles after a month" and "把手用三次就松" describe the same problem. Top positive and negative themes, ranked by frequency.
Actionable Insights.
Every report ends with a P0–P3 priority list mapped to evidence — no more "we read 800 reviews and learned nothing actionable." Ready to plug into Listing edits, ads copy, or supply chain decisions.
Amazon review data is hard and expensive to get.
Front-end scraping is brittle, multi-marketplace data is fragmented, and reading 800 reviews by hand teaches you nothing structured. Here's what most sellers hit.
800 reviews = 80 pages
Amazon front-end shows 10 reviews per page. Copy-paste 5 competitors costs 10 hours minimum.
Patterns don't surface
Read 150 1-star reviews and you'll know "buyers complain a lot" — but not which complaint is industry-wide vs one-off.
Marketplaces are siloed
Want German buyer feedback before launching DE? US-marketplace view doesn't show DE reviews. 14 markets, 14 manual scrapes.
Data → action is missing
You scraped, you read, you noted themes. Now what? Most tools stop at the data — Luckee stops at the decision.
Full scope of what Review Analysis covers.
Data collection, AI semantic clustering, double-document reporting, and competitive enforcement — structured for sellers who need answers, not raw exports.
Dual-source scraping. 14 marketplaces. 3 depth modes.
- Dual-Source Coverage Primary + Secondary API cross-verified. 85–100% of written reviews captured, deduplicated, normalized.
- 14 Amazon Marketplaces US, UK, DE, FR, JP, CA, IN, ES, IT, MX, AE, AU, BR, SA. US runs deep mode (500–700 reviews); others standard (100).
- 3 Scrape Modes Quick Preview (100), Standard (500), Deep (500–700 default). Pick speed vs depth per use case.
- Full Metadata Date, Verified Purchase, helpful votes, image/video flags, variant attributes (color, size), VP percentage.
Semantic clustering. Not keyword search.
- Theme Clustering AI understands "handle wobbles" + "把手很松" + "handle broke" describe the same defect. Top 5 positive + Top 5 negative themes auto-ranked.
- Frequency & Severity Each theme: occurrence %, severity (High/Medium/Low), representative quotes, industry-wide vs single-competitor flag.
- 3★ Signal Analysis 3-star reviews are the richest — buyers say both pros and cons. Luckee specifically surfaces these structural concerns.
- Multi-Language Auto-processes DE/JP/FR/ES reviews — extracts themes regardless of source language.
Two documents. Decisions, not exports.
- reviews-data.md Full corpus organized by star rating — date, variant, VP status, original text. Audit-ready.
- reviews-summary.md Basic stats + theme tables + 3★ signals + time trends + Key Findings + P0–P3 actionable recommendations with evidence.
- Time Trend Detection Auto-flags crisis windows (e.g. "May–Jul 2025: 6/7 negatives clustered = likely batch defect") and recovery periods.
- Markdown-Native Drops into Notion, Linear, Slack, or any editor. No PDF export friction.
Catch competitor review fraud. With evidence.
- 15-Signal Cross-Check Review rate spike, concentration window, 5★ ratio anomaly, content similarity, VP ratio drop — multi-signal validation, not single-metric guess.
- RED / AMBER / GREEN Verdict Composite scoring across all 15 signals. RED = report-ready; AMBER = monitor; GREEN = clean.
- Auto-Generated Complaint Letter Amazon-format draft: violating ASIN, evidence timeline, data comparison, representative quotes, violation type. Review and submit.
From ASIN to decision in 5 steps.
Input one ASIN. Select marketplace + mode. Luckee handles the rest — scraping, dedup, clustering, reporting.
Primary API
Max-mode scrape for primary review source.
Secondary API
Cross-source scrape for marketplace coverage.
Dedup & Merge
Normalize, dedupe, unify metadata schema.
Generate Data
Full review corpus by star rating (Deliverable 1).
Generate Summary
Themes + trends + actionable recommendations (Deliverable 2).
See it on a real ASIN.
A standing desk launched in 2023. 37 reviews collected. A May–July 2025 motor failure crisis surfaced. 5-month recovery confirmed. P0–P3 recommendations attached.
Standing Desk Analysis
37 written reviews · Aug 2023 – Feb 2026 · US Marketplace · Dual-source @ ~100% coverage
Full review corpus organized by star rating — 37 reviews with date, variant, VP status, helpful votes, and original text. Audit-ready, drops into Notion or Linear.
Basic stats + Top themes + 3★ signals + time trends + 6 Key Findings + 8 P0–P3 actionable recommendations (e.g. "Audit 40-inch White variant" / "Fix after-sales SLA" / "Redesign two-piece desktop joint").
Where teams put Review Analysis to work.
Competitor full-corpus scrape
Input 5 ASINs. Minutes, not hours. 800-review products hit ~95–100% coverage. Auto-deduplicated, star-grouped, metadata-complete.
Negative-theme prioritization
AI semantic clustering. Top 5 negative themes auto-ranked with frequency, severity, representative quotes, and industry-wide vs competitor-specific flags.
Multi-marketplace pre-launch
14 marketplaces in one tool. Pull DE / JP / FR reviews; auto-translates and clusters themes regardless of language.
Reviews → action gap
P0–P3 priority list with mapped evidence. Each recommendation cites which themes / quotes / time trends justify the priority.
Suspect competitor review fraud
15-signal cross-validation surfaces the anomaly with quantified evidence. Auto-generates an Amazon-format complaint letter — you review, you submit.
Trend-based opportunity scouting
Run Review Analysis across competitor set. Industry-wide negative themes with persistence over time = real product opportunity, not noise.
Why this isn't something Helium 10 already does.
Side-by-side, where Luckee Review Analysis pulls ahead of manual scraping and existing seller tools.
| Manual Scraping | Helium 10 / Jungle Scout | Luckee Review Analysis | |
|---|---|---|---|
| Coverage per ASIN | ~10–30% (copy fatigue) | 50–70% (single source) | 85–100% (dual-source) |
| Time per 800 reviews | ~2 hours | ~15 min export | Minutes, automated |
| Theme clustering | Manual highlight | Keyword frequency | AI semantic clustering |
| 3★ structural analysis | — | — | ✓ Dedicated |
| Time trend detection | — | Basic monthly chart | ✓ Crisis + recovery windows |
| 14 marketplaces | Per-market manual | US-heavy | ✓ All 14, one tool |
| P0–P3 action list | — | — | ✓ Evidence-mapped |
| Compliance / complaint letter | — | — | ✓ Auto-drafted |
Before you try it.
What kind of coverage can I expect on a high-volume ASIN?
Dual-source scraping covers 85–100% of written reviews (note: Amazon's total review count includes star-only votes with no text — these aren't scrape-able by anyone). On the B0BLCBRBVZ case above we captured 37 written reviews ≈ 100% coverage.
Which marketplaces are supported?
US, UK, DE, FR, JP, CA, IN, ES, IT, MX, AE, AU, BR, SA — 14 total. US supports Deep mode (500–700 reviews). Other markets currently support Standard (100 reviews per scrape).
How fast is a scrape?
A 500-review Deep scrape on US typically completes in 2–5 minutes including dedup, clustering, and both deliverable documents. A 5-competitor batch is parallelized.
What does the AI clustering actually do that keyword search doesn't?
It understands that "handle wobbles after a month", "把手用三次就松", and "handle broke after 4 weeks" describe the same durability defect. Keyword search would treat these as three separate items. Semantic clustering merges them, weights frequency, and surfaces severity.
Can I integrate this into our existing tooling?
Both deliverables are Markdown. Drop into Notion, Linear issues, Slack, GitHub README, or any text editor. API access for the data layer is available on request.
How does the compliance / fraud-detection module work?
15-signal cross-validation runs on demand per ASIN: review rate, concentration window, 5★ ratio, content similarity, VP percentage, etc. A composite RED/AMBER/GREEN verdict tells you whether to report. If RED, you get an Amazon-format complaint letter draft with the evidence baked in.
Stop reading reviews. Start acting on them.
Run Review Analysis on one of your ASINs free. See both deliverables in under 5 minutes.