The shelf life field is on 3.1% of records, and it holds only four numbers

About a 5 minute read.

"Is this bottle still good?" is probably the most-asked and least-answerable question. The database has a shelf_life field that looks purpose-built for it. I went to see what it actually holds.

First: it is nearly empty

	Records	Share of database
Has a shelf life	1,303	3.1%
Has storage conditions	1,232	2.9%

1,303 out of 42,225.

This time "nearly empty" matters less than usual — 1,251 of those 1,303 have suppliers (96%). The field tracks things that are actually on sale.

Second: it holds only four numbers

Months	Records
24	561
12	397
6	186
36	111
18	21
48	10
everything else (3, 8, 9, 10, 25, 60, 64)	17

Four round numbers cover 1,255 records — 96.3%.

This was not measured. It was sorted. Nobody determined that something lasts 25 months — there is exactly one 25-month record, and it reads more like a typo. What seems to have happened is closer to: someone put materials into four piles labelled half a year, a year, two years, three years.

Third: the number is a floor, not a deadline

Every entry's full string is the same sentence:

24.00 month(s) or longer if stored properly.

"Or longer." That promises a minimum. It does not declare an end.

It is the opposite of what you assume. A date on food means "past this, don't eat it." This number means "at least until here, it still counts." Not one record tells you when to throw anything away.

Fourth: the numbers do not track oxidation risk

I expected this field to roughly mirror how easily something oxidises — citrus short, musks long. Pulling the 100 records whose odor_type is citrus:

Months	Records
12	38
24	34
6	16
18	6
36	6

Citrus is scattered across every bucket.

And the cleanest contradiction is this one:

Material	Shelf life	Suppliers
Sweet orange peel oil	36 months	99
Sweet orange peel oil, folded (terpene-stripped)	12 months	122

Folded is the version with the terpenes removed. The terpenes are the half that oxidises; taking them out should make it keep better. It gets a third of the plain oil's number.

The rest show no pattern either: grapefruit oil 12 months, lemon oil 12, d-limonene 12, citral 24, citronella oil 24.

I do not know how these numbers were set. But they are not a ranking of oxidation rate.

The usable information is in a different field

The storage field is far more useful than shelf_life, and simpler — there are only four strings in the whole database:

String	Records
Cool, dry place, tightly sealed, protected from heat and light	995
Refrigerate in tightly sealed containers	145
Cool, dry place … store under nitrogen	50
Store under nitrogen	42

"Refrigerate" appears on 145 records. That is the strongest instruction in the field, and it is independent of the month count — the refrigerate group holds 102 records at 12 months and 26 at 6, but also 7 at 24. The most-supplied of them is geranyl acetate: 130 suppliers, 24 months, refrigerate.

"Nitrogen" appears on 92 records (7.5% of those with a storage field). The top odour type in that group is citrus (15), which matches intuition.

So the practical reading order is: use storage to decide how to keep it, then read shelf_life as a very coarse reference. Not the other way round.

And the thing to worry about is not "going off"

The problem with an old bottle is not that it smells bad — you can smell that yourself. The problem is that oxidation creates sensitisers that were not in the bottle when you bought it.

Matura and colleagues tested 1,511 consecutive dermatitis patients across six European dermatology centres in 2005, in Contact Dermatitis. Their premise is stated plainly:

Previous studies have shown that limonene, linalool and caryophyllene are not allergenic themselves but readily form allergenic products on air-exposure.

The results: 1.3% of patients reacted to oxidised linalool, 1.1% to its hydroperoxide fraction, about 0.5% to oxidised caryophyllene. Of those reacting to the oxidised terpenes, 58% had fragrance-related contact allergy or a history of adverse reactions to fragrance.

Those three materials' shelf lives in the database: limonene 12 months, linalool 24, β-caryophyllene 24. And of the three, only linalool is flagged for nitrogen.

That is why the month count is not enough. The number is about whether something is still saleable. Matura's paper is about the fact that six months after you opened it, the thing in the bottle is not what you bought.

What to actually do

• Follow the storage field where it exists, especially those 145 refrigerate records.
• For the 92 nitrogen records, fill the bottle when you decant. Reduce headspace. The 100 mL bottle you have used 10 mL of is the dangerous one.
• Treat the month count as a rough reference, not an expiry date. Its own wording is "or longer."
• For the classes that oxidise — citrus, terpenes, aldehydes — write your own opening date. The database will not count for you.
• Re-evaluate on a new lot. I made the same point in the birch tar article for a different reason.

What this doesn't establish

• I do not know who set those month figures or on what basis. The field gives no source and no test method.
• "Four round numbers means sorted, not measured" is an inference from the shape of the distribution, not from any documentation.
• I have not explained the sweet orange peel oil 36 versus folded 12 contradiction. I can only confirm that both fields read that way.
• "Citrus is scattered across every bucket" is an observation at the level of correlation; I ran no statistical test.
• Matura 2005 tested deliberately oxidised terpenes, not bottles left for N months. The air exposure was a laboratory condition; I have not established how it maps onto the bottle on your bench.
• I ran no storage experiment. This is field statistics and one paper.
• Both shelf_life and storage have about 3% coverage, so every distribution above describes those ~1,300 records, not the database.