Note 03
What grading actually is
Grading is the work of turning a variable heap into a described product with a stated tolerance, and it is a skilled judgement rather than a sorting chore.
What is being decided when material is graded?
Grading is often imagined as tidying: separating the big pieces from the small ones. That is one operation within it, and not the one that carries the risk.
What grading actually decides is what the enterprise is prepared to say about a consignment, and how wrong it is willing to be. A grade is a claim. It says this material falls within these bounds. Everything downstream, including the price, the buyer’s process settings and the enterprise’s credibility, rests on that claim holding for the load that arrives rather than for the handful that was inspected.
That reframing matters, because it explains why grading cannot be casually delegated or automated away at this scale. The judgement being exercised is about representativeness and variation, not about size.
What is being measured?
Four properties do most of the work.
Size fraction. The distribution of particle sizes in the material, usually established by passing a sample through screens of decreasing aperture and weighing what is retained at each stage. A buyer’s process is built around a size range. Material that is too coarse may need crushing the buyer has not budgeted for. Material with too many fines can create handling and dust problems, and in some processes fines behave differently in the reaction itself.
Composition. What the material is made of, and in what proportions. At this scale composition is assessed by experienced visual and physical inspection, supported by laboratory analysis where a buyer requires it. Visual assessment is genuinely informative in a known geology, where a practised grader has handled thousands of samples from the same formations. It is not a substitute for assay where an assay is required, and it should not be presented as one.
Moisture. How much water the material is carrying at the point of sampling. This matters twice over. It affects handling and process behaviour, and it affects weight, which is what the seller is paid on. A consignment sold wet and weighed wet transfers the cost of that water to the buyer, which a buyer notices quickly. Moisture is why the sampling moment has to be recorded along with the figure.
Contamination. What is in the material that should not be. Soil, vegetation, plastic, tramp metal, or material from a different source mixed in at the stockpile. This is usually the easiest to fix and the most damaging to leave.
How is a sample taken, and why is method the whole game?
A grade derived from a badly taken sample is worse than no grade, because it carries unearned confidence.
The problem is segregation. Granular material does not stay mixed. Move it, pour it, or let it sit, and the coarse fraction migrates in one direction while the fines settle in another. The top of a heap is not the middle. The material near the mouth of a sack is not the material at its base. A sample taken from wherever is convenient is a sample of the most accessible fraction, which is systematically not the average.
The practices that address this are unglamorous and old.
- Take increments from multiple points across the lot rather than one grab from one place.
- Take them from moving material where possible, or from depth rather than the surface.
- Combine increments into a composite, then reduce that composite down by halving repeatedly rather than by scooping from it, so that reduction does not reintroduce the bias sampling just removed.
- Record where, when and by whom the sample was taken, and retain a portion of it.
That retained portion is the part that inexperienced operations skip. It is what allows a dispute to be resolved rather than argued. If the buyer’s analysis disagrees with the seller’s, and both parties can go back to a retained split of the same sample, the disagreement becomes a technical question. Without it, the disagreement is about whose word to accept, and the smaller party loses that by default.
What is an acceptable variation?
No consignment matches its specification exactly. The useful question is what range the seller commits to and what happens outside it.
A stated grade should therefore come with a tolerance and a consequence. Within the range, the load is accepted at the agreed price. Outside it, something defined happens: a price adjustment, a rejection, a re-grade at the seller’s cost. Naming this in advance is what separates a supply relationship from a series of arguments.
Sellers new to this tend to state tolerances that are too tight, because a narrow range sounds like quality. A range you cannot consistently hit is not a quality claim, it is a rejection schedule. The stronger position is a range that is honestly achievable, stated plainly, and met every time.
Why does this work stay in the county?
There is a version of formalisation in which the material leaves the area raw and the grading, the value and the skill all accumulate somewhere else. The county exports sacks and imports nothing.
Grading is the natural point of resistance to that, for a practical reason rather than a sentimental one. It depends on knowing the local material. A grader who has worked the same formations for years reads a sample faster and more accurately than an outsider with better equipment and no context. That knowledge does not travel, and it cannot be bought in a week.
Keeping grading local therefore keeps the highest-skilled and highest-judgement part of the line where the material is, along with the wage that attaches to it. It also puts the people who know the material in the position of making the claim about it, which is where that decision belongs.
Where this enterprise currently stands
Screening and assessment are carried out by the women associated with the enterprise, at the recovery sites, using hand screens and physical inspection. Sampling method and retention are part of what we are formalising rather than something we can present as long-established practice.
We do not publish our screen apertures or the size fractions we offer, because we are establishing both during the pilot. A published fraction is a commitment to produce it consistently, and the point of a pilot is to find out what can be produced consistently before committing to it. Nor do we publish a tolerance, for the same reason and with more force: a tolerance is a commercial promise, and one made before the evidence exists is a rejection schedule waiting to happen.
We do not currently hold laboratory analysis for the material. Where a buyer requires assay, that would be arranged and the result would travel with the consignment.
The short version
Grading turns a variable heap into a described product with a stated tolerance. Its difficulty lies in sampling, because segregation makes convenient samples systematically unrepresentative. The safeguards are increments, halving, and a retained split. The skill is local, which is the strongest argument for the work staying where the material is.