RankquantRQ

Rankquant normalized consumer-product ratings

1,352,809 consumer products scored with the Rankquant per-reviewer normalization method, derived from an Amazon customer reviews review corpus of 4,721,328 reviewers. Each reviewer is re-centred against their own rating distribution before aggregation, so a percentile reflects relative standing rather than a raw star average. Rows carry the z-normalized percentile and a sample-size-adjusted percentile, each also recomputed within the row's peer cohort. 1,215,077 of 1,352,809 rows carry a normalized percentile; the rest carry no rating figure at all — no source-computed average is substituted. Published in parts rather than as one file — a cross-category index file holding the top 150,000 rows by score1, 14 per-category index files, 256 slug-sharded detail files covering every row — so no single download holds the whole catalog. JSON, CC BY 4.0.

Read this before using the file

Coverage is partial by design, and this is the one vertical where that matters. The Amazon review capture is one page deep per listing, so most reviewers are seen exactly once and can supply no personal baseline. Rows without enough reviewer overlap carry no rating at all — not a Rankquant percentile, and not the retailer’s star average as a stand-in. Filter on zRanked before treating score1 or score3 as present. No column in this dataset carries a source-computed aggregate: the retailer’s star average and rating count are collected as scrape diagnostics and are never published. This vertical also ships no raw-average lens, so there is no score2 or n2 column.

Download

This catalog is published in parts. It holds 1,352,809 consumer products; the single ranked file below holds 150,000 of them — the top of the ranking by score1. The 14 per-category files carry the rest of the browse tier, and the detail files carry every row.

Provenance

Provenance and parameters for Rankquant normalized consumer-product ratings
Rows in the catalog1,352,809 consumer products
Rows in index.json150,000 — the top of the ranking by score1, capped
Detail files256, together covering all 1,352,809 rows
Review sourceAmazon customer reviews
Source reviewers4,721,328
Rows carrying a normalized score1,215,077 of 1,352,809
Score schemareviewer-z-r3-mean-v2
Shrinkage constant K53
Data generated2026-08-07

Columns

Every row of index.json and of the per-category files carries these fields. Percentiles run 0–100, where 100 is the top of the catalog.

Column definitions for Rankquant normalized consumer-product ratings
FieldDefinition
slugURL-safe identifier; resolves to https://rankquant.com/products/<slug>/.
nameDisplay title of the consumer product.
score10–100. Percentile of the mean per-reviewer z-score, taken across reviewers with two or more ratings — including those whose ratings are all identical, whose dispersion is imputed from the pooled source dispersion. The headline Rankquant score: each reviewer is re-centred on their own scale before aggregation.
score1Cohortscore1 recomputed within the row's peer cohort instead of against the whole catalog.
score30–100. mean_z × n/(n + 53) — the same reviewer-normalized mean behind score1, pulled toward the corpus average in proportion to how thin the sample is — then ranked against the distribution of the unshrunk means, so the pull is visible in the percentile. A row keeps the fraction n/(n+53) of its measured distance from the mean. Shrinkage arithmetic — no model is fitted and nothing is inferred.
score3Cohortscore3 recomputed within the row's peer cohort.
n1Count of calibrated reviewers behind score1 and score3 — reviewers with two or more ratings, including those whose ratings are all identical. A row with 2 or more is indexed; a row with exactly 1 is scored but not indexed.
asinAmazon Standard Identification Number.
categoryRetail category the product is listed under.
nReviewsReviews Rankquant collected and scored for this product. Our sample size — not the retailer’s rating total.
zRankedBoolean. True when the row has enough reviewer overlap to carry a real normalized percentile. False rows carry NO rating of any kind — not a Rankquant score and not a borrowed source figure.
zIndexableBoolean. True when the row clears the stricter threshold for a standalone indexable page.

The detail files carry every column above plus per-row diagnostics the browse files leave out — among them rawAvg, the plain mean of the star ratings Rankquant collected on a 1–5 scale, unnormalized: a sample statistic, not a rank, and not the retailer’s average. Those columns are absent from index.json and from the per-category files.

Method

Each reviewer is re-centred against their own rating distribution before anything is aggregated, so a percentile reports relative standing inside a peer set rather than a raw star average. The full derivation, including every constant, is published at /methodology/. The sample-size adjustment behind score3 is shrinkage arithmetic — a row keeps the fraction n/(n+53) of its measured distance from the corpus average. Nothing is fitted or inferred.

Licence and reuse

Licensed under CC BY 4.0. Republish, cite, or remix with attribution to Rankquant and a link to this page. Questions about bulk access or a different export format: [email protected].