# Dupedex

> Dupedex records which merchants publicly state that their fragrance is inspired by which
> designer original, and verifies each statement against the merchant's own product page.
> Every record carries the evidence URL, the date it was read and the reviewer. We do not
> rate similarity and we do not rank merchants. No merchant listed here pays us: links to the
> dupes carry no affiliate tag. Links to the original on Amazon are paid links, carry
> rel="sponsored", and are disclosed at page level — the site footer and /affiliate-disclosure.
> As an Amazon Associate, Dupedex earns from qualifying purchases.

Dataset version: 1.9.3+2026-08-23 (schema 1.7.0, generated Aug 23, 2026)
Licence: CC BY 4.0 — attribution required, commercial use permitted.

## Scale

- 3807 verified claim records
- 2551 fragrances with at least one verified claim
- 463 brands
- 7 merchants tracked, 6 of them read in full
- 5387 merchant product pages read; 2013 of them read and not promoted, each with a reason code
- 7614 ledger events
- Human review pass rate for this release: 89.0%

## Canonical surfaces

- [Home and search](https://dupedex.com/): the finder over all 2551 records.
- [Fragrance index](https://dupedex.com/fragrance): the browse surface, shelved by what the copies start at.
- [Brand directory](https://dupedex.com/fragrance/brands): all 463 houses, A to Z, with the record count for each.
- [Brand hubs](https://dupedex.com/fragrance/brands/tom-ford): one page per brand, 463 of them.
- [Entity sheets](https://dupedex.com/fragrance/): `/fragrance/{brand}-{name}` — the authoritative page for one fragrance: every verified claim naming it, prices, sizes, cost per millilitre, listed stock, and the merchant's own wording quoted with a link to the page it was read from.
- [Merchants](https://dupedex.com/merchants): coverage stated as pages read over pages published, plus a page per merchant splitting claim honesty from fulfilment facts.
- [Dataset and coverage](https://dupedex.com/dataset/coverage): version, counts, per-merchant progress, checksums, licence and citation format.
- [Changes](https://dupedex.com/changes): the event log in plain English.

## Guides

Two reference explainers, written from primary sources with every citation linked. They are
prose, not ledger records: they carry editorial dates rather than a dataset version.

- [What “inspired by” means in fragrance](https://dupedex.com/guides/what-does-inspired-by-mean): It is a seller’s comparison, dated and attributable — not a lab result, a licence, or a statement from the house being named. 9 linked sources; last edited Aug 23, 2026.
- [How fragrance clone houses work](https://dupedex.com/guides/how-clone-houses-work): Where the liquid comes from, who is responsible for it, and which parts of the price a merchant’s own explanation can and cannot account for. 16 linked sources; last edited Aug 23, 2026.

## Machine exits

- [dataset.csv](https://dupedex.com/dataset/dataset.csv): one row per verified claim.
- [dataset.jsonl](https://dupedex.com/dataset/dataset.jsonl): one JSON object per verified claim, all fields.
- [non-claims.jsonl](https://dupedex.com/dataset/non-claims.jsonl): pages read and not promoted, with reason codes.
- [manifest.json](https://dupedex.com/dataset/manifest.json): version, counts, licence, sha256 per file.
- [changes.atom](https://dupedex.com/changes.atom): the most recent ledger events as a feed.
- [finder-index.json](https://dupedex.com/finder-index.json): the search index the site's own finder uses.

## How to cite

Dupedex dataset v1.9.3+2026-08-23, checked 2026-08-23, https://dupedex.com/dataset/coverage

Attribution is the licence condition: name Dupedex and link the version you used. Quote the
dataset version, because releases are immutable — a correction produces the next version
rather than editing a published file.

## What the data does and does not assert

- A record asserts that a named merchant published a named inspiration statement on a named
  URL on a named date, and that a reviewer matched it to a fragrance down to brand, flanker
  and edition.
- A record does not assert that the two fragrances smell alike. No similarity score exists in
  this dataset and none will be added.
- Price, listed stock and catalogue presence are three separate fields, each carrying the date
  it was read. Anything older than 45 days is reported as unknown rather than as current.
- `unknown` is a legal value throughout and means "we looked and found no published answer".

## Crawling

Public pages are open to search and AI crawlers, including training user agents. There is no
paywall, no account and no rate limit. Please cite the dataset version.

## Contact

Corrections: https://dupedex.com/editorial-policy