ScentPrism

What this is, and what is ours

ScentPrism is a standard, and it is upstream of this site. The vocabulary, the facets, the families and the note artwork all arrive here in a versioned release and are applied wholesale. We do not edit them locally: anything we changed would be overwritten by the next release, and the two would disagree in the meantime for reasons nobody could trace.

What ScentVerdict owns is the attribution - which notes a given perfume has, and in what proportion. Everything on the pages below describes the standard we read from, not a thing we can change by hand. Where our own research suggests the standard is wrong, the deliverable is evidence sent upstream.

Why it matters if you are just here about a perfume

Three things follow from having one vocabulary rather than several. Every note means the same thing on every page, so "amber" is not quietly two different smells in two places. Every classification can be traced back to the notes that produced it, rather than asserted. And two perfumes can be compared on the same coordinates, which is what makes a similarity score mean anything at all.

The map

The 10 families gather all 34 facets between them. The facets are the coordinate system; the families are how it is read. Both are drawn here from the same definition the wheel on every perfume page uses, so this diagram cannot drift out of step with the product. Every family and every facet has its own page.

Explore the standard

Each part is its own page, so it can be read, linked and argued with on its own terms.

• How it is built The five layers, from CAS numbers up to the families a shopper browses
• The 34 facets What each one covers, where it stops, and the substances that carry it
• The 10 families How the 34 are gathered, and why a family value is a max and not a mean
• What we have measured The numbers behind the model, and what they do not prove
• The note vocabulary, as JSON Every published note as JSON, with its coordinates and how many perfumes use it

Why there are no accords here

Until August 2026 our perfume pages carried a bar chart of accords: woody, fresh spicy, amber. We took them off. They remain standard across the category, so it is worth being direct about why they are absent from a model that claims to describe the same thing.

An accord looks like a second opinion. It is not. The accord data used across the fragrance web is produced by mapping a perfume's notes to family labels and weighting them by user votes on those notes. No chemical analysis is involved. It is a derivation of the notes, plus a popularity signal - so holding both put two answers to one question on the same page, produced by two methods from a single input, with nothing to adjudicate between them. When we traced our own data, 202,737 of 248,566 accord rows (81.6%) had been copied from one external source of exactly that kind, and 103 of 161 accords already existed as notes under the same name.

It was visible on the page, not just in the data. A perfume's summary would read "a patchouli-led wear" from an imported accord label while the wheel a few hundred pixels below read "Floral, dominant" from the perfume's own notes. Same perfume, same screen. That sentence now reads "a floral-led wear", because it and the wheel are computed from the same numbers.

The case for treating accords as irreducible rested on emergence - the idea that a blend produces something you cannot recover from its parts. In July 2026 a Monell Chemical Senses Center team quantified this across 432 mixtures of 144 odorants with a trained panel and found nearly all were explained by a linear average of their components' quality profiles, including mixtures previously reported to show emergence.

The concept is not gone from the site. The accords guide still explains what an accord is and how to read one in a fragrance description, which is the question people actually arrive with.

Where this is going

ScentPrism is being built as a versioned standard: facet definitions, a note vocabulary with coordinates, mappings to other family systems, and the method behind all three, each released under a version number so a change can be pointed at rather than argued about.

Fragrance has one widely licensed classification and no open coordinate layer. We are building the coordinate layer because we need one; how it is published is a decision that has not been made yet, and this page will say so when it is.