Full Overview or Star Buckets? Choosing the Right Scrape Mode

The first decision when submitting a task: full overview, or star buckets. The two modes are mutually exclusive — one per task. Choose wrong and you either burn credits on data you'll never open, or economize your way past the signal you actually needed. Here's the framework for choosing.

Why two modes exist

It starts with the rating distribution: it tells you what happened — the five-star share, the one-star rate. But it can't tell you why. The "why" lives in the review text inside each band: what one-star reviewers are angry about, what five-star reviewers love. Full overview and star buckets are the two answers to breadth versus depth.

Full overview: the fast pulse

The full overview samples evenly across all star bands, about 10 reviews per page, with the page count (1–10) set by you and auto-capped by the product's actual review volume. It answers "how is this product doing overall": a quick health check during product screening, a first read on a new product's reputation, a fast comparison pass over competitors. Full-band coverage, low cost, fast — at the price of limited depth in any single band.

Star buckets: targeted depth

Bucket mode scrapes only the star bands you choose, each with its own 1–10 page count, in any combination — only 1★ and 2★ to mine complaint root causes, or only 5★ to analyze what delights buyers. It answers "what is this specific band actually saying": one-star root-cause mining, five-star motivation analysis, or a targeted check when one band's share looks anomalous. Up to roughly 500 reviews — far deeper than the overview.

How it's billed

Both modes are priced identically, by the page: a fixed probe fee per product (covering the product info probe — price, title, rating), plus a per-page credit rate × total pages — the overview page count, or the sum of every bucket's pages. Two guardrails protect you: page counts are capped by the product's real review volume, so you never pay for pages that don't exist; and any bucket the probe confirms empty is refunded in full for its pages (the probe fee itself is never refunded). Exact prices live on the pricing page.

Three worked decisions

• New seller, screening products: run a 3–5 page overview first — healthy rating structure, real demand — and only then go deeper; if it looks promising, follow up with a low-star bucket pass on complaint root causes.
• Established product, one-star rate spiking: skip the overview. Open 1★ and 2★ buckets at ~5 pages each and mine the shared root causes, separating product issues from expectation mismatches.
• Competitive teardown: overview the competitor for the shape of their reputation, then bucket only their one-star band — their pain-point list is your differentiation list.

One rule of thumb when unsure: run a small page count first, read the rating distribution, then decide whether to add buckets — the data tells you what to capture next.