Refined or unrefined: what you pay for in a cold-pressed oil
Two listings, same oil, same supplier. One says refined; the other says unrefined, cold pressed, virgin, has a nicer photograph, and costs two or three times as much per kilo. You are trying to work out whether the difference is a real material difference or a story, and nobody will give you a straight answer, because the people selling the expensive one have an interest and the people selling the cheap one have a different interest.
There is a straight answer, and it is quantitative. Refining removes measurable amounts of specific things, and how much it removes has been measured. Whether you should pay for them depends entirely on whether your formula uses them.
This piece is part of our series built around plant oils by fatty-acid group. It is about the one or two per cent of an oil that is not a triglyceride at all — the unsaponifiable fraction — because that is what refining takes and that is what the price gap is mostly about. How fast an oil oxidises (iodine value, peroxide value, induction period) belongs to the companion article on oxidative stability of oils and is not repeated here.
What the word on the label actually means
Start with a surprise: the clearest definitions of the words on that listing are in a food standard, not a cosmetic one.
Codex CXS 210-1999 §2.2 defines them:
Virgin oils are obtained, without altering the nature of the oil, by mechanical procedures, e.g. expelling or pressing, and the application of heat only. Cold pressed oils are obtained, without altering the oil, by mechanical procedures only, e.g. expelling or pressing, without the application of heat.
That is the whole of it. Cold pressed means no heat was applied during pressing. It does not mean the oil is fresher, better filtered, richer in anything, or handled more carefully after pressing — and it is defined for foodstuffs, in a standard whose scope is oils "presented in a state for human consumption."
On the British side, the closest thing to a definition of "natural" for a cosmetic ingredient is a standard rather than a law: BS ISO 16128-1:2016, published on 31 March 2016, which defines natural ingredients as "cosmetic ingredients obtained only from plants, animals, micro-organisms or minerals" and organic ingredients as natural ingredients from organic farming or wild harvesting. Read its own scope statement before you lean on it: the standard explicitly does not address "product communication (e.g. claims and labelling), human safety, environmental safety and socio-economic considerations", nor regulatory requirements. It is a shared vocabulary for suppliers, not permission to say anything.
At the bench: write into your raw-material specification the things that are actually defined and measurable — species, refining state, the unsaponifiable figure and the method that produced it — and stop treating "cold pressed" or "natural" as a quality grade. They are descriptions of a process and a vocabulary, respectively.
Unsaponifiable matter: a number defined by its method
Saponify a plant oil and nearly all of it turns into soap. What is left, extractable with a solvent, is the unsaponifiable matter: the material that was never an ester of a fatty acid, plus the unsaponifiable halves released from esters.
The definition is operational rather than chemical, and this matters more than it sounds. Both relevant Codex standards name the same three methods — ISO 3596:2000, ISO 18609:2000 or AOCS Ca 6b-53. The principle is what the name implies: reflux the oil with potassium hydroxide in ethanol, extract with a solvent (diethyl ether in ISO 3596, hexane in ISO 18609), wash the extract free of soap and alkali, evaporate the solvent, weigh the residue. What is reported is the mass of that residue, as g/kg or as a percentage.
Two consequences follow, and both cost people money.
The figure is a gravimetric total, not a composition. Two oils with identical unsaponifiable percentages can contain entirely different things.
Two different solvents are not guaranteed to agree on the same sample. A certificate reporting unsaponifiable matter without naming the method is reporting half a number.
And the Codex figures themselves are ceilings, not typical contents. In Table 2 of CXS 210-1999 the unsaponifiable row is a single value per oil against which an authentic crude oil is checked. CXS 33-1981 is explicit, heading the column "Maximum level".
| Oil (crude, unless stated) | Unsaponifiable matter limit | Standard |
|---|---|---|
| Palm kernel oil, arachis (groundnut) oil, high-oleic safflower | 10 g/kg each | CXS 210-1999, Table 2 |
| Palm oil, babassu oil | 12 g/kg each | CXS 210-1999, Table 2 |
| Coconut, cottonseed, sunflowerseed, soya bean, mustardseed, safflowerseed | 15 g/kg each | CXS 210-1999, Table 2 |
| Virgin, refined and blended olive oil | 15 g/kg (maximum) | CXS 33-1981 |
| Grapeseed, rapeseed, low-erucic rapeseed, sesameseed | 20 g/kg each | CXS 210-1999, Table 2 |
| Maize oil | 28 g/kg | CXS 210-1999, Table 2 |
| Olive-pomace oils | 30 g/kg (maximum) | CXS 33-1981 |
| Rice bran oil | ≤ 65 g/kg | CXS 210-1999, Table 2 |
The spread is the point. Rice bran oil is allowed more than six times the unsaponifiable matter of groundnut oil — and rice bran oil is the one sold on the strength of that fraction.
At the bench: ask the supplier which ISO method produced the unsaponifiable figure on the certificate, and record it on the same line in your notes. If they cannot say, the number is not comparable with anyone else's and should not be used to justify a price difference.
Tocopherols: the oil's own antioxidant, and its fingerprint
Tocopherols are natural antioxidants that contribute to an oil's oxidative stability alongside its fatty-acid composition, processing history and storage. There are four principal homologues — alpha, beta, gamma and delta — differing in the number and position of methyl groups on the chromanol ring. Appendix Table 4 of CXS 210-1999 tabulates them for crude oils from authentic samples; unlike the Table 2 ceilings, these are observed ranges given for identification.
| Crude oil | Alpha-tocopherol (mg/kg) | Gamma-tocopherol (mg/kg) | Total tocopherols and tocotrienols (mg/kg) |
|---|---|---|---|
| Sunflowerseed | 403–935 | ND–34 | 440–1520 |
| Soya bean | 9–352 | 89–2307 | 600–3370 |
| Maize | 23–573 | 268–2468 | 330–3720 |
| Sesameseed | ND–3.3 | 521–983 | 330–1010 |
| Rapeseed (low erucic) | 100–386 | 189–753 | 430–2680 |
| Rice bran | 49–583 | ND–212 | 191–2349 |
| Palm | 4–193 | ND–526 | 150–1500 |
| Grapeseed | 16–38 | ND–73 | 240–410 |
| Coconut | ND–17 | ND–14 | ND–50 |
ND means not detectable under the conditions the table represents. Read it across rather than down and the pattern jumps out: the isomer profile is as characteristic of an oil as its fatty acids. Sunflower is an alpha-tocopherol oil. Soya bean and maize are gamma-tocopherol oils. Sesame carries almost no alpha at all. Palm and rice bran carry much of their vitamin E as tocotrienols instead — the same table gives palm 4–336 mg/kg alpha-tocotrienol and 14–710 mg/kg gamma-tocotrienol, and rice bran up to 627 and 790 mg/kg of the same two. Coconut has the lowest total in the table, ND–50 mg/kg, which is one reason nobody buys coconut oil for its antioxidants.
The ranges are wide because the content is agronomic rather than fixed. Maranz and Wiesman (2004) measured tocopherols by HPLC in 102 shea butter samples from 11 African countries and found total tocopherol from 29 to 805 micrograms per gram, mean 220, with alpha-tocopherol the principal form at an average 64 per cent. Climate accounted for much of the spread: hot, dry N'Djamena in Chad averaged 414 µg/g alpha-tocopherol, cool highland northern Uganda 29 µg/g. A fourteen-fold difference in the same commodity, and no fatty-acid figure on a certificate would show it.
What tocopherol content does not do is let you predict shelf life. It is one input among several — fatty-acid profile, pro-oxidant metals, peroxides already present, packaging, headspace, storage temperature. A high natural tocopherol figure is a reason to expect a certain robustness, not a substitute for a stability study.
At the bench: if you are buying an unrefined oil specifically for its own vitamin E, ask for the tocopherol figure on the lot, not the species range. If the supplier does not measure it, you are paying for an average.
Phytosterols: most of the fraction, by mass
For most seed oils the largest single component of the unsaponifiable matter is not tocopherol but sterol. Appendix Table 3 of CXS 210-1999 gives total sterols for authentic crude oils in mg/kg:
| Oil | Total sterols (mg/kg) |
|---|---|
| Rice bran | 10,500–31,000 |
| Maize | 7,000–22,100 |
| Sesameseed | 4,500–19,000 |
| Rapeseed (low erucic) | 4,500–11,300 |
| Sunflowerseed | 2,400–5,000 |
| Soya bean | 1,800–4,500 |
| Coconut | 400–1,200 |
| Palm | 300–700 |
Codex reports total sterols of 10.5–31 g/kg for authentic crude rice bran oils alongside an unsaponifiable ceiling of 65 g/kg. Those are independent figures from independent tables and should not be divided into each other to produce a "typical sterol proportion."
The same standard tabulates the principal desmethylsterols as percentages of total sterols, and the proportions are diagnostic. Beta-sitosterol dominates nearly everywhere — 76.0–87.1 per cent in cottonseed, 57.7–61.9 in sesameseed, 54.8–66.6 in maize, 45.1–57.9 in low-erucic rapeseed. Campesterol and stigmasterol follow in proportions that identify the oil: rapeseed is unusual for campesterol at 24.7–38.6 per cent, soya bean for stigmasterol at 14.9–19.1 per cent.
Olive is policed more tightly still. CXS 33-1981 sets campesterol at not above 4.0 per cent of total sterols and a minimum total sterol content of 1,000 mg/kg for virgin olive oils, refined olive oil and olive oil. Those are authenticity criteria — a sterol profile is how an adulterated olive oil gets caught — but they also tell a formulator what an olive oil's fraction is made of. (Editions of this standard change; check the current text before quoting a limit in a document of your own.)
On the skin side, the evidence a cosmetic formulator can honestly lean on is narrower than the marketing round phytosterols suggests. Puglia and Bonina in the Journal of Cosmetic Science (2008) applied a formulation containing soybean phytosterols to human volunteers whose stratum corneum had been disrupted by tape stripping, and followed barrier recovery non-invasively by reflectance spectrophotometry of methyl-nicotinate-induced erythema. Three days after stripping, the sites given the phytosterol formulation showed an appreciable recovery of barrier function compared with sites given the vehicle alone. That is a single small in-vivo study on a barrier model in intact skin, and it says nothing about any finished product. It also licenses no claim to treat or affect a disease or condition — a phytosterol figure on a certificate is not in itself a claim about anything, and it should not become one on a label.
At the bench: sterols are the mass of the fraction, which is why the unsaponifiable figure moves when the sterols move. They are not the reason most formulas specify an unrefined oil; if that is your reason, name the study you are relying on and keep it in your file.
Squalene and squalane: one letter, two materials
Squalene is a triterpene hydrocarbon, C₃₀H₅₀, with six carbon-carbon double bonds and no oxygen at all. It is a biosynthetic intermediate on the route to sterols and triterpenes, which is why it appears in the unsaponifiable fraction of oils whose plants make plenty of those. Squalane is the same skeleton with every double bond hydrogenated away: C₃₀H₆₂, fully saturated, without the sites that make squalene oxidise. One letter, and a wholly different stability profile.
Olive is the classic plant source, and the best quantification available is unusually clean. Hernández and colleagues in the Journal of Agricultural and Food Chemistry (2023) quantified squalene by gas chromatography in virgin olive oils from a core collection of 36 cultivars — all grown in the same orchard under the same conditions, harvested at the same ripening index of 2.5, and extracted under identical conditions, so that genotype was very nearly the only variable left. The mean was 4.52 mg/g of oil and the range 1.27 to 11.83 mg/g: 0.13 to 1.18 per cent of the oil, a nine-fold spread. They also observed a continuous decrease in squalene during ripening in oils from 'Picual' and 'Arbequina'.
Three readings follow for a formulator. A virgin olive oil brings a real but variable quantity of squalene, at a level closer to a few tenths of a per cent than to the figures sometimes quoted for it. That squalene is not a stabiliser — it is another component with six double bonds in it. And cosmetic squalane, whether from olive or from sugarcane fermentation, is the hydrogenated derivative, chosen precisely because a saturated hydrocarbon does not do what squalene does on a shelf.
At the bench: if you are buying a virgin olive oil for slip you are buying a fraction of a per cent of a variable component, and if you want that slip reliably you want squalane as an ingredient in its own right. Swapping one name for the other because they look alike is among the more expensive mistakes with these materials.
Triterpene alcohols, and why shea is the outlier
Shea butter is the fat where the unsaponifiable fraction stops being a rounding error. Di Vincenzo and colleagues (2005) determined triacylglycerol, fatty-acid and polycyclic triterpene composition in 150 shea samples from Mali, Burkina Faso, Nigeria and Uganda, and found country means for acetyl and cinnamyl polycyclic triterpenes ranging from 3.69 to 12.57 per cent, the highest in Nigerian provenances, with high variability in all three classes of compound. The triterpene alcohols involved are alpha- and beta-amyrin, lupeol and butyrospermol, largely esterified with acetic and cinnamic acids.
Those results identify shea as compositionally unusual. Variation in this fraction may well help explain why two lots with similar stearic and oleic figures set and feel differently — but the cited study measured composition, not those properties, and the distinction is worth keeping. What it means for dosing shea against cocoa and mango is worked through in which butter makes the balm you want.
Avocado, discussed among the oleic oils, is the other oil in this series with a notably large fraction — and it makes a different point. Santana and colleagues (Food Chemistry, 2019) expeller-pressed Hass avocado pulp under a range of drying and peeling conditions and reported unsaponifiable matter of 2.48–2.99 g/100 g in the oils from unpeeled microwave-dried pulp, with alpha-tocopherol of 11.6–21.0 mg/100 g and induction periods of 54.2–83.6 hours. Pressing conditions, not just the fruit, decided what ended up in the oil.
At the bench: when your formula depends on something in this fraction — a shea whose triterpenes give it its body, an unrefined avocado brought in for its own tocopherols — specify the material at that level of detail and re-check it when the lot or the origin changes. The fatty-acid profile will not tell you it has moved.
What refining actually takes out — with numbers
Refining is designed to remove free fatty acids, phosphatides, colour, odour and contaminants. The unsaponifiable fraction is not the target, but much of it leaves with them. The clearest sign of this is commercial: the deodoriser distillate drawn off in the final step of vegetable oil refining is itself traded as a source of tocopherols and phytosterols. What the refinery sells as a by-product is what left your oil.
The clearest quantification is Rhazi, Depeint and Ayerdi Gotor in Molecules (2022), who followed minor components of sunflower oil through six steps of an industrial chemical-and-physical refining process. Across the full sequence:
| Component | Loss across refining |
|---|---|
| Carotenoids | 98.6 % |
| Squalene | 45.0 % |
| Phytosterols | 19.5 % |
| Tocopherols | 8.5 % |
They also observed that the largest reductions fell on the compounds that most affect the visual appearance of the oil — waxes, carotenoids, chlorophylls — while reduction was limited for compounds with no impact on organoleptic quality. Refining is selective, and selective in favour of the tocopherols: the antioxidant survives better than the pigment or the squalene.
Better is not untouched, and the size of the effect depends on the oil and the process. Asbbane and colleagues in Scientific Reports (2024) compared argan oil that had been bleached and deodorised — physical refining, explicitly in the context of preparing a cosmetic ingredient — against oil from the same initial batch, over twelve weeks of accelerated storage at 60 °C. Physical refining improved the oil's initial quality, removing free fatty acids (by up to 30 per cent) along with primary and secondary oxidation products — and produced what the authors describe as a dramatic decrease in oxidative stability, with tocopherol losses of up to 94 per cent after twelve weeks under those conditions.
The two studies are not in conflict; they measure different things. One is an immediate loss on processing. The other is a loss of protection through subsequent storage. Together they say: refined oil starts cleaner and ages worse.
Codex records the consequence in the plainest possible way. CXS 33-1981 §4.1 permits no additives at all in virgin olive oils; §4.2 permits alpha-tocopherols to be added to refined olive oil, olive oil and the pomace grades "to restore natural tocopherol lost in the refining process", with the concentration of alpha-tocopherol in the final product not to exceed 200 mg/kg. A food standard that has to authorise putting the antioxidant back is telling you exactly what refining removed.
At the bench: refined is the right buy when the fraction is not doing anything in your formula — a wash-off product, a soap, a base oil in a formula that carries its own antioxidant, or anything where the colour and odour of the unrefined grade are a nuisance. Unrefined is the right buy when a named component of that fraction is the reason the oil is in the formula. "Unrefined because it's better" is not a specification and will not survive a costing review.
What to do with the number at the bench
| Question | What unsaponifiable data can support | What it cannot |
|---|---|---|
| Is this oil what the label says? | A figure outside the Codex limit for that oil, or a sterol profile outside the tabulated ranges, is a reason to query the lot | A compliant figure does not by itself establish authenticity, origin or grade |
| Refined or unrefined? | Refining measurably reduces carotenoids, squalene and sterols, and to a lesser degree tocopherols — 98.6 / 45.0 / 19.5 / 8.5 per cent across six steps on sunflower | Those percentages do not transfer to another oil or another refinery |
| Why do two lots behave differently? | Provenance and processing move the fraction a long way — 29 to 805 µg/g total tocopherol in shea, 1.27 to 11.83 mg/g squalene across olive cultivars | It does not predict the direction or size of a difference in a given formulation |
| How stable will the product be? | Nothing on its own; it is one input to a stability programme | It is not a shelf-life figure, and no natural antioxidant content removes the need to test the finished product |
| Can I say "natural" on the pack? | BS ISO 16128-1 gives you a shared technical vocabulary with your supplier | That standard states in its own scope that it does not address claims and labelling — what you may write is a separate question, treated in what a UK label can say about an oil |
Two habits pay for themselves. Ask which ISO method produced the unsaponifiable figure, because the number is defined by its method. And when a formula depends on a specific component of that fraction, buy the oil on that component and re-check it when the lot or the origin changes — because the fatty-acid profile will not tell you it has moved, and the price will not either.
Composition informs the choice. It promises nothing about the finished product, which has to be validated on each formulation and each production batch.
Sources
- Codex Alimentarius. Standard for Named Vegetable Oils, CXS 210-1999 (adopted 1999, amended 2015): §2.2 definitions; §5 methods of analysis (ISO 3596:2000 / ISO 18609:2000 / AOCS Ca 6b-53); Table 2 unsaponifiable matter; Appendix Table 3 desmethylsterols and total sterols; Appendix Table 4 tocopherols and tocotrienols. FAO PDF — consulted 22 September 2026.
- Codex Alimentarius. Standard for Olive Oils and Olive Pomace Oils, CXS 33-1981 (amended 2015): sterol composition, minimum total sterols, unsaponifiable matter maximum levels, §4.1 and §4.2 food additives. FAO PDF — consulted 22 September 2026.
- ISO 3596:2000, Animal and vegetable fats and oils — Determination of unsaponifiable matter — Method using diethyl ether extraction, and ISO 18609:2000, …Method using hexane extraction. Cited as reference methods in CXS 210-1999 and CXS 33-1981.
- BS ISO 16128-1:2016, Guidelines on technical definitions and criteria for natural and organic cosmetic ingredients and products — Definitions for ingredients. Published 31 March 2016. BSI Knowledge
- Maranz S., Wiesman Z. Influence of climate on the tocopherol content of shea butter. Journal of Agricultural and Food Chemistry 52(10), 2934–2937, 2004. PubMed 15137838
- Di Vincenzo D. et al. Regional variation in shea butter lipid and triterpene composition in four African countries. Journal of Agricultural and Food Chemistry 53(19), 7473–7479, 2005. PubMed 16159175
- Hernández M. L. et al. Functional characterization of four olive squalene synthases with respect to the squalene content of the virgin olive oil. Journal of Agricultural and Food Chemistry 71(42), 15701–15712, 2023. PubMed 37815987
- Puglia C., Bonina F. In vivo spectrophotometric evaluation of skin barrier recovery after topical application of soybean phytosterols. Journal of Cosmetic Science 59(3), 217–224, 2008. PubMed 18528589
- Santana I. et al. Hass avocado (Persea americana Mill.) oil enriched in phenolic compounds and tocopherols by expeller-pressing the unpeeled microwave dried fruit. Food Chemistry 286, 354–361, 2019. PubMed 30827618
- Rhazi L., Depeint F., Ayerdi Gotor A. Loss in the intrinsic quality and the antioxidant activity of sunflower (Helianthus annuus L.) oil during an industrial refining process. Molecules 27(3), 916, 2022. PubMed 35164180
- Asbbane A. et al. Effects of physical refining process on quality and stability of argan oil (Argania spinosa (L.) Skeels). Scientific Reports 14(1), 23045, 2024. PubMed 39367085
