Beginner perfumer · 93
The evaluation concentration field holds five values, and this time the ladder is real
· 12 min read
My last few articles have been about fields that do not mean what they say: shelf life is four round numbers that do not track oxidation risk, storage arrives with the record's source, a FEMA number is not an identifier. This one goes the other way. Of the 8,298 records with an evaluation entry, 7,620 (91.8%) state a concentration, and there are only 14 distinct values — five decade rungs (100%, 10%, 1%, 0.1%, 0.01%) cover 99.6%. It looked like another sorted-not-measured field. But crossed against strength rank, the correspondence is perfectly monotonic: at 100% the strongest rank is 0.3% of records, at 10% it is 28.3%, at 1% it is 82.4%, at 0.1% it is 86.5%, and at 0.01% it is 100%.
About a 4 minute read.
My last few articles have been about pulling fields apart: shelf life is four round numbers that do not track oxidation, storage arrives with the record's source, a FEMA number is not a material identifier.
This one goes the other way.
Another field with only a handful of values
Entries in the evaluation field (odor_descriptions) usually look like this:
at 1.00% in dipropylene glycol. fruity floral berry plum…
Of the 8,298 records with an evaluation field, 7,620 (91.8%) state a concentration. And there are only 14 distinct values:
| Concentration | Records | Share |
|---|---|---|
| 100% | 5,297 | 69.5% |
| 10% | 1,068 | 14.0% |
| 0.1% | 588 | 7.7% |
| 1% | 532 | 7.0% |
| 0.01% | 107 | 1.4% |
| the other nine values | 28 | 0.4% |
Five decade rungs cover 99.6%.
At this point I expected the same story as before — someone sorting materials into a few piles that correspond to nothing real, the way shelf life does.
But this one lines up
Crossing concentration against strength_rank (1 weakest to 3 strongest):
| Evaluation concentration | Records with a rank | Rank 1 | Rank 2 | Rank 3 (strongest) |
|---|---|---|---|---|
| 100% | 1,332 | 17.9% | 81.8% | 0.3% |
| 10% | 520 | 0.8% | 71.0% | 28.3% |
| 1% | 193 | 0.0% | 17.6% | 82.4% |
| 0.1% | 37 | 0.0% | 13.5% | 86.5% |
| 0.01% | 3 | 0.0% | 0.0% | 100% |
Perfectly monotonic. The strongest rank climbs from 0.3% to 100% without once turning back. The weakest rank appears only at 100% and 10%, and disappears entirely below.
This ladder was measured, not sorted.
And what it means is plain: how far down you have to dilute this before you can stand to smell it.
What is on the bottom rung
The odour types of the 107 records at 0.01%:
| Records | Odour type |
|---|---|
| 29 | sulfurous |
| 9 | nutty |
| 8 | alliaceous |
| 8 | meaty |
| 7 | green |
Sulfur, allium, meat, nuts — the families with the lowest thresholds in flavour chemistry.
The best-supplied among them: 2,3,5-trimethyl pyrazine (99 suppliers), maple furanone (67), coffee difuran (54), allyl isothiocyanate (53), garlic oil (53), methyl mercaptan (50), caramel furanone (41), phenol (28), pyridine (23).
This rung is not a corner of obscure chemicals. These are heavily used materials.
One observation running the other way
Look at the second column above: the lower the concentration, the fewer records carry a strength rank at all.
At 100%, 1,332 records have one (25%). At 0.01%, only 3 do (2.8%).
The strongest materials are the least likely to have their strength recorded.
My guess is that for these, the strength field has run out of resolution — calling something "high" says nothing about a material you have to dilute ten-thousand-fold to smell. The concentration itself has become the measure of strength.
The two records that say 1000%
One of the 14 values is physically impossible: 1000%.
| #4136 bergamot oil terpeneless (11 suppliers) | "at 1000.00 %. sweet citrus linalyl acetate" — Luebke 1998 |
| #5351 (0 suppliers) | "at 1000.00 %. muguet, green, and fatty notes" |
A concentration cannot exceed 100%. Most likely "100.00" with an extra zero.
This kind of error is easy to catch — because this field has a physical ceiling. Most fields do not.
In practice
- This number tells you directly how to start. 100% means smell it neat; 0.01% means three tenfold dilutions before you begin.
- It is more useful than the text
odor_strengthfield. Strength is filled on 25% down to 2.8% of records; concentration on 91.8%. - The same material described at different concentrations reads differently, and the evaluation field often holds both. Align the concentrations before comparing.
- This is the most honest field I have measured so far. Shelf life, storage,
potential_uses, the toxicity field's units — all of those need discounting. This one does not.
What this doesn't establish
- Concentrations are pulled from the evaluation text by a regular expression (
at N %), so records written differently are missed and 91.8% is a floor. - Strength rank coverage is low, and only 3 records at the bottom rung — that row's 100% is three out of three, far too small a sample to lean on. The trend rests on the middle rungs (520, 193 and 37 records).
- The monotonicity is descriptive; I ran no statistical test.
- Strength rank and evaluation concentration may not be independent judgments — the same evaluator very likely chose a dilution first and wrote the strength afterwards. If so, the agreement between these two fields is circular rather than corroborating. This is the biggest weakness here, and nothing in the data rules it out.
- I judged the two 1000% entries to be typos without checking any source.
- I evaluated nothing myself.