How it is built
Five layers. Each one is worked out from the layer below it. Nothing is imported alongside. That is what makes two answers to the same question impossible, rather than something we have to police forever.
The layers
Read down and you go from physical reality to marketing language. The two picked out in the middle are the layers ScentPrism adds. The category already has the other three in some form.
Real substances with a CAS number. A perfumer can buy them. Scoped by the IFRA Transparency List. This is the layer that makes everything above it checkable against sources outside our own database.
"Bergamot". "Cashmere wood". "Ocean breeze". A controlled vocabulary, where every term is typed by how it relates to a material. Some are a single substance. Some are dominated by one. Some are a reconstruction with no natural source. Some denote nothing you can buy. We record which is which, and publish the distinction rather than flattening it.
The coordinate system. Every note carries a position on all 34 at once, and a spread on each saying how much that position varies between compositions. The spread is our error bar. We publish it.
The facets gathered into 10 groups a person can read. This is what the wheel on a perfume page draws, and what a shopper browses by.
Other classification systems, expressed as mappings into the same space. The intent is to publish each with a measured agreement score, so you can see where a mapping holds and where it breaks. What is drawn below is the shape of the problem, not the mappings themselves.
Physical reality Marketing language
Three things this category keeps confusing
Telling them apart is most of the work. It is also most of the value.
A real substance with a CAS number. Cashmeran. You can buy it.
The word a brand prints on the box. "Cashmere wood". Marketing language.
One axis of the space. Creamy wood. One of 34 things we measure.
A material smells of facets. A note is made real by materials. A perfume lists notes. The link from a note to its material is the piece nobody else publishes. It is also what lets you check a facet value against a source outside our database.
Every other system becomes a view
Fragrance has several ways of sorting perfumes into families, and they disagree. One system calls a perfume "Fresh Aromatic". Another calls the same perfume "Fougere". Publishing one more would just add one more thing to disagree with. So we publish the layer underneath instead, and map the others onto it. Four of them are below. Follow any one to its own page and check the counts at source.
Four systems, four answers
One coordinate
Your labels, our coordinates
CIELAB did not beat Pantone. It became the space Pantone is defined in, and Pantone now publishes Lab values for its chips. Unicode did not win by being a better Latin-1. It won by publishing mappings back to every old encoding, so moving to it never cost you what you already had.
In both cases the label layer was fashion. The coordinate layer was infrastructure. Fragrance has four label layers and no coordinate layer.