Beginner perfumer · 73
A third of the records sharing a CAS number are filed under different odour types
· 13 min read
A CAS number gets used as a guarantee that two things are the same substance. For naturals it is not. Records with an odour type in this database spread across 4,536 CAS numbers, of which 357 carry more than one record — and in 127 of those 357 groups (35.6%) the records are filed under different odour types. It is not random either: lavender shares CAS 8000-28-0, where the oil is floral and the absolute and concrete are herbal; lemon shares 8008-56-8, where sixteen of seventeen records are citrus and "lemon oil distilled" is floral; labdanum and cistus share 8016-26-0, where labdanum absolute is amber and cistus twig/leaf absolute is woody. One plant, different extractions, one number.
About 4 minutes.
When two people on a forum need to check they are talking about the same material, the standard advice is "compare the CAS numbers".
For a single molecule that works well. For naturals it fails, and the failure rate can be measured.
Measure it first
Pulling every record that has both a CAS and an odor_type, they spread across 4,536 distinct CAS numbers.
Of those, 357 CAS numbers carry more than one record.
And of those 357 groups, the records are filed under different odour types in 127 — 35.6%.
More than a third.
What the disagreement looks like
Sorted by total supplier count, the largest groups:
Lavender (CAS 8000-28-0, 273 suppliers)
| Record | Odour type | Suppliers |
|---|---|---|
| Lavender oil | floral | 99 |
| Lavender oil France | floral | 55 |
| Lavender oil Bulgaria | floral | 50 |
| Lavender absolute France | herbal | 39 |
| Lavender concrete | herbal | 10 |
The oil is floral type; the absolute and the concrete are herbal.
Lemon (CAS 8008-56-8, 311 suppliers)
Seventeen records, sixteen of them citrus, with one exception:
| Record | Odour type | Suppliers |
|---|---|---|
| Lemon oil | citrus | 74 |
| Lemon oil distilled | floral | 24 |
| Lemon oil Italy | citrus | 47 |
"Lemon oil distilled" is filed as floral, and it has 24 suppliers.
Labdanum / cistus (CAS 8016-26-0, 217 suppliers)
| Record | Odour type | Suppliers |
|---|---|---|
| Labdanum absolute | amber | 58 |
| Cistus twig/leaf absolute | woody | 51 |
| Cistus twig/leaf oil | woody | 49 |
| Labdanum oil | amber | 18 |
| Cistus leaf oil | balsamic | 3 |
Three odour types — amber, woody and balsamic — under one CAS, and two different plant names in the record titles (labdanum and cistus are both Cistus ladaniferus).
Other groups with the same pattern:
- Geranium (8000-46-2): oil is floral, absolute is green, concrete is herbal
- Basil (8015-73-4): anisic and licorice mixed together, absolute is herbal
- Orris (8002-73-1): concrete and absolute floral, CO2 extract waxy, tincture woody
- Ylang / cananga (68606-83-7): the flower oils are all floral, cananga leaf oil is spicy
Why this happens
A CAS number for a natural is assigned to the plant, not to the product.
CAS 8000-28-0 is defined by a description along the lines of "lavender oil, Lavandula angustifolia", and it covers extracts of that plant. But a steam-distilled oil, a solvent-extracted concrete and the absolute washed out of that concrete can differ substantially in composition, because the extraction method decides which molecules get through.
- Steam distillation carries the volatile fraction
- Solvent extraction brings waxes, pigments and high-boiling material along
- CO2 extraction is another line again
So "the same plant" is true and "the same thing" is not.
(A worked case of one CAS holding several products is Ylang ylang oil: four different products sharing one CAS; this article measures that across the whole database. Batch variation within a natural is How much does a natural vary.)
What this means in practice
One: a CAS match is not enough to confirm you are discussing the same material. Ask for the extraction too — oil, absolute, concrete, tincture, CO2.
Two: searching by odour type will miss things. Search herbal and you miss lavender oil; search floral and you miss lavender absolute — same plant. Search citrus and you miss a distilled lemon oil with 24 suppliers.
Three: changing extraction is changing material, even at the same CAS. Swapping lavender oil for lavender absolute is not a change of supplier; the database itself files them under different odour types.
Four: the odour type field is an editorial judgement. Of that 35.6%, I cannot separate how much is a real difference in smell from how much is different labels applied at different times. Both are possible, and that alone limits what this field can be used for.
What to do
- Report the extraction alongside the material. "Lavender absolute (France)" is far more useful than "lavender oil, CAS 8000-28-0".
- Do not rely on a single odour type when searching. Products from one plant may be scattered across two or three types.
- Re-evaluate when you change extraction. Treat it as a new material.
- Synthetic single molecules are not affected. One molecule, one CAS, and comparing CAS numbers works.
What this doesn't establish
- This is an audit of field consistency, not a measurement of odour. I measured how the database labels things, not how different they smell.
- The denominator for "35.6%" is the 357 groups where one CAS carries several records that all have an odour type. Records without one are excluded, and changing the filter changes the number.
- Disagreement does not mean error. An oil and an absolute genuinely can smell different, in which case different odour types are correct.
- I did not separate real differences from editorial inconsistency. That is the biggest limit here.
- I did not look up compositional analysis for any group. "Extraction decides which molecules get through" is general chemistry, not my measurement of these groups.
- CAS definitions for naturals are loose to begin with, and registration and usage practices differ between suppliers. I did not check CAS's own definitions record by record.