Two things determine whether a kernel purchase works: what the kernel looked like when it left the packing house, and what happened to it between then and the factory door. Grading covers the first. Oxidation control covers the second. Buyers who specify the first carefully and leave the second to good intentions are the ones who open a sound-looking carton and find a rancid product.
Where the grading rules come from
There is no EU regulation that grades nut kernels. The frameworks the trade uses are conventions, and they bind only where a contract names them.
The UNECE standards for dry and dried produce are the most widely referenced. They are commodity-specific: DDP-02 for walnut kernels, DDP-06 for almond kernels, and parallel standards for hazelnut and other kernels. Each sets out presentation forms, quality classes, sizing provisions and defect definitions with tolerances by class. They are trade conventions, not Union law, and a contract that does not name the standard and the class has not invoked them.
The International Nut and Dried Fruit Council quality guidance sits alongside as the industry’s own reference for sampling practice, analytical approach and defect vocabulary. Again: convention, invoked by naming it.
What is law is the safety layer underneath. Aflatoxin maximum levels sit in Commission Regulation (EU) 2023/915, with sampling and analysis methods in Commission Implementing Regulation (EU) 2023/2782. Pesticide residues sit in Regulation (EC) No 396/2005. None of these are grade parameters and a higher grade carries no safety advantage.
Colour is the dominant price lever on walnut
For walnut kernel, colour outranks every other parameter commercially. The trade grades run from extra light through light, light amber and amber to darker classifications, and the price spread between the top and the middle of that range is large enough to dominate the economics of a contract.
Two points are easy to get wrong. First, colour must be judged against a physical reference under standardised lighting. Colour perception under warehouse fluorescents and under daylight are different, and a colour dispute settled by two people looking at kernels on a table is settled arbitrarily. Second, colour is not fixed. A light kernel darkens with age, heat and oxygen exposure, so an extra light grade purchased and then stored badly becomes a light amber kernel without anyone having done anything dishonest. That makes colour a joint grading and storage parameter, and it is the reason a colour claim on arrival needs a defined measurement point.
Hazelnut colour works differently. The commercial split there is between natural, blanched and roasted, and within blanched the parameter that matters is residual skin – the pale kernel with visible skin remnants is the one that shows up as speckle in a praline or a white chocolate filling.
Physical form: halves, pieces and the fragment question
Presentation form is a mechanical specification. Halves, quarters, large pieces, small pieces and meal are separate articles with separate yields and separate prices, and the buyer’s job is to specify the minimum content of the named form plus a tolerance for everything below it.
The most common failure in this area is specifying halves without specifying the minimum halves percentage. Cracking always breaks some kernel; no operation delivers one hundred per cent halves. A specification that names the form without naming the minimum is a specification the seller will interpret in their own favour, entirely legitimately.
Shell fragments are the separate and more serious question. A shell fragment is hard, sharp and immediately noticeable to a consumer, and it generates complaints out of all proportion to its frequency. Writing absent is unenforceable, as it is for any physical contaminant. What works is a stated count per unit weight combined with a described removal system: gravity separation, optical sorting with shell-specific detection, and end-of-line metal and X-ray inspection where the value justifies it. Ask which of those the supplier actually runs, in sequence.
Sizing and count
Kernel sizing is expressed either as a count per unit weight or as a screen fraction, depending on the commodity and the standard invoked. Larger kernels go to visible, whole-kernel applications and carry a premium; smaller kernels and pieces go to paste, chopping and inclusion applications where size is invisible.
The same logic applies as elsewhere in this document: pay for size only where size is seen. A confectionery plant chopping kernel into a bar is paying a premium for nothing if it buys on a whole-kernel size grade.
Oxidation: reading the two numbers together
Kernel stability is a lipid chemistry problem and it is measured by two numbers that only mean something in combination.
Peroxide value, determined by the iodometric method in ISO 3960, measures primary oxidation products. Its critical property is that it is not monotonic: peroxides form and then decompose into secondary products, so an old, badly stored kernel can return a low peroxide value because the reaction has already moved past its peak. A single peroxide reading with no date and no temperature history is close to uninformative.
Free fatty acid content, by ISO 660, measures hydrolytic rather than oxidative breakdown. It rises steadily rather than peaking, which makes it the stable half of the picture, and it reports mainly on the condition of the raw material and on the interval between harvest and processing.
Read together the two separate a raw material problem from a storage problem, which is exactly the distinction a rejection conversation turns on. High free fatty acids with low peroxide value points at the crop and the drying. Low free fatty acids with high peroxide value points at storage and handling after processing. Both high points at a product that is simply old.
A sensory check completes the picture and is faster than either analysis. Rancidity is detectable by an untrained palate well before it becomes a formal specification failure, and a taste of every incoming lot is the cheapest quality control in the category.
The oxidation drivers, in order of practical importance
- **Temperature.** Reaction rate rises steeply with temperature; a warm warehouse shortens shelf life far more than most buyers assume. Chilled storage roughly multiplies practical shelf life on high-fat kernels.
- **Oxygen.** Vacuum packing, nitrogen flushing or an oxygen barrier liner all work. Plain polyethylene in a carton does not.
- **Light.** Relevant only for clear packaging, which means retail rather than bulk.
- **Surface area.** Chopped, sliced and milled kernel oxidises far faster than whole. A kernel meal has a fraction of the practical life of a half.
- **Skin removal.** The seed coat carries natural antioxidants. Blanched kernels are less stable than natural ones, which is a real cost of the blanching decision.
- **Prior heat treatment.** Roasting accelerates subsequent oxidation, which is why roasted kernel shelf life is shorter than raw and why roasting should be as close to use as the supply chain allows.
Why the storage clause matters more than the specification line
Because oxidation is driven by conditions rather than by the kernel itself, a peroxide value limit on the specification sheet protects the buyer only at the measurement point named in the contract. If that point is the seller’s warehouse at despatch, a compliant lot can arrive non-compliant after a hot transit and the buyer has no claim.
A functional contract therefore names four things: the measurement point for the oxidation panel, the packaging barrier specification, the temperature ceiling during transport and storage, and the responsibility split for temperature excursions. High polyunsaturated seeds – hemp, flax and hulled sunflower in particular – need this treatment more urgently than tree nuts, because their oxidation clocks run considerably faster.
A specification checklist
- Name the grading standard and the class, or the standard is not invoked
- State the colour grade with a physical reference and a lighting condition
- State the presentation form with a minimum percentage and a tolerance
- Give shell fragments a count per unit weight and ask for the removal sequence
- Specify size only where the kernel is visible in the finished product
- Specify moisture and water activity together, with methods
- Specify peroxide value and free fatty acids together, with the measurement point
- Specify the barrier packaging and the transport temperature ceiling
- Keep the aflatoxin file separate: it is a legal obligation, not a grade parameter