Where this data comes from

1. Where clone relationships come from — two tiers

Tier one, editorial (222 clones). The public clone index at clonespreadsheet.com names a clone and an original, with a stated match percentage and price snapshots. We matched each side to a fragrance record by brand and name, then hand-checked the pairs where the naming didn't line up. Those percentages are editorial claims, not measurements, and they are labelled as claims everywhere they appear.

Tier two, discovered from wearers (977 clone pairs). Every fragrance page in the source corpus carries a "smells like this" carousel with agree/disagree vote counts. Where an Arabic or clone-house fragrance is repeatedly voted as smelling like a designer or niche original, we record that as a clone relationship at scale — no editor required.

• Strong (455): at least 85% of voters agree, on at least 25 votes.
• Likely (522): at least 75% agree, on at least 15 votes.
• Cross-check: of the editorial pairs that also carry votes, the median agreement is 86% and 81% clear this threshold — so tier two recovers the same kind of relationship the editorial list records.
• A cloned fragrance can point at more than one original; extra ones appear as also compared to.

2. Why both numbers sit next to each other

Editorial claims and community votes are independent, and they disagree often enough to matter. When a clone has both, we show both: the claimed match percentage and the share of wearers who voted it really does smell like the original.

Where the two diverge, the page says contested instead of quietly dropping the clone. Where there is no editorial claim at all, the community vote is the claim, and it is labelled that way.

3. Fragrance data

Notes (with per-note prominence weights), accords, official gender, release year, community ratings, longevity, projection, season and day/night votes come from a Fragrantica corpus snapshot published on Kaggle (140,228 perfumes, CC BY-NC-SA 4.0).

• 6,889 records in this site's curated set: every perfume from 26 Arabic and clone houses, plus designers and niche houses above a popularity threshold.
• 96% of records carry longevity and projection vote data; 99% carry a usable note pyramid.
• Ratings are shown as-is with vote counts, so you can judge how thin the sample is.

Bottle photography: the product shots published next to each clone on the clone index (originals only — the index illustrates each clone with a picture of the fragrance it copies). Where there is no photo, the card is generated from that fragrance's dominant accords. Personal, non-commercial project; imagery is credited and will be removed on request.

Short write-ups marked Editor review (100 of them) are written for this site from the structured data — accords, notes, season, performance, price gap and the community vote — not copied from any review database. Prices are snapshots captured on the date printed next to them and are not live.

4. What this site does not do

• It does not sell anything and it does not use affiliate links. If you buy, buy from whoever you trust.
• It does not claim a clone is identical. A 96% match is a claim about how a fragrance reads to people, not a lab result.
• It does not publish bottle photography. The colour cards are generated from each fragrance's dominant accords.

5. Licensing

Fragrance data is used under CC BY-NC-SA 4.0 — non-commercial, share-alike, attribution. That is why this site carries no advertising and no monetised links. Derived data on this site is shared under the same terms.

Corrections, missed clones and wrong pairings: the highest-value contribution is telling us a clone is mis-paired. Every fragrance page carries both signals precisely so the errors are visible.