Amazon Review Scraping: Full Overview vs Star Buckets

How cloud scraping works, mechanically

Every task is one ASIN from one of nine marketplaces (US, Germany, UK, Australia, Mexico, India, Egypt, UAE, Canada — detected from the link you paste). Submission returns immediately; the scrape runs in a queue and progress is visible per task. Reviews arrive roughly 10 per page, and a single task can pull up to 10 pages per page-carrying row.

What you get back is the raw review pool: star rating, text, and metadata per review, downloadable as CSV or JSON whether or not you ever generate a report. The quick-start guide walks the full flow from submission to report.

The two modes, side by side

The modes are mutually exclusive — one task runs either an overview row or a set of star buckets, never both. Pick per task; nothing is locked in.

Both modes bill identically per page — 5 credits per page plus 1 base credit per ASIN — so mode choice is about data shape, not price. The full cost logic is in credits and billing.

Full overview: when breadth is the point

Overview mode scrapes the newest reviews regardless of star — which is also its honest weakness: on a 4.6-star product, the sample skews toward 4s and 5s, and a handful of recent 1-stars may not make it in. That is fine when the question is "what does the review pool look like overall" — sentiment structure, positioning, first-look diagnostics on a new ASIN, or feeding the report a general sample.

• Use it when: validating a niche before entering, checking a supplier's product, first analysis of your own listing.
• Skip it when: you already know the rating and need to know why — a 10-page overview of a heavily 5-star product tells you little about the 1-star root causes.

Star buckets: star-balanced depth

Bucket mode scrapes each star band you select as its own sample with its own page count. Because bands don't overlap, three pages of 1-star plus two pages of 5-star guarantees both voices are present in proportion you chose — the low-star root causes cannot be diluted by volume. This is the mode the report engine was built around: root-cause clustering shines when the complaint pool is dense.

Two mechanics worth knowing:

• Probe row: bucket tasks start with a cheap probe that reads the product's real star distribution first, so bucket depths are set against reality — and the base fee rides on this zero-page row rather than being charged per bucket.
• Empty-bucket refund: if the probe shows a band has no reviews (a 4.9-star product has almost no 1-stars), that bucket is falsified and its page fees are refunded automatically. The probe's own base fee stays — the probe did real work.

Max depth is five buckets × 10 pages × ~10 reviews — roughly 500 reviews — which is the ceiling of a single task's coverage.

Choosing: three worked examples

• Entering a new niche — start with a 5-page overview on each of two or three candidate ASINs, read the sentiment structure, then decide whether a deeper bucket run is warranted. Cost so far: 26 credits per ASIN.
• Your own product at 4.3 stars — go straight to buckets: 3–4 pages of 1-star and 2-star (root causes), 2 pages of 5-star (promises to keep). Roughly 41–51 credits for a diagnosis you can hand to a factory.
• Competitor teardown at 1.9 stars — overview is enough: the pool is already dense with complaints; buckets would add cost without new signal.

For the longer decision framework — sampling theory, dilution math, and more examples — the blog post full overview vs star buckets is the deep dive. To see what a report built on star-balanced data looks like, open any diagnostic example report.

Scraping modes FAQ

Continue reading