khamush.com Lifestyle | Motivation | Poems
Two rail cars pull into the plant carrying the same mineral, from the same supplier, with assay sheets that look almost identical. One runs through the process without complaint. The other pushes yields down, drifts the furnace, and has the shift supervisor on the phone before lunch.
Nobody switched the material. The numbers matched. And yet the plant floor knows something the paperwork doesn’t.
This is the assay gap, and it’s older than most of the equipment it frustrates. A certificate of analysis is a summary of a sample, not a description of a load. The distance between those two things is where a lot of quiet money gets lost.
The Certificate Describes a Sample, Not the Load
A mineral shipment is rarely uniform in the way a liquid is uniform. Fragments differ in size, density, and mineral content, and they segregate during loading, transport, and unloading. The scoop or auger that pulled the lab sample captured a slice of that variability, not the average of it.
Formal sampling theory has a name for the ceiling this puts on precision. Work by Pierre Gy, developed over decades of industrial practice, shows that the error is a function of fragment size, mineralogical composition, and how the material was gathered. Two honest samples from the same pile can, and often do, disagree.
Bias Is the Quiet Part of the Number
Random variation is the error operators expect. Bias is the one they don’t. A sampler mounted where fines collect, a splitter that favors coarser lumps, a probe that never reaches the bottom of a bag: each of these tilts every result in the same direction, quietly, for years.
An analysis of Gy’s framework lays out just how much of the total error budget belongs to preparation and handling rather than the lab bench. The assay isn’t wrong. The path the material took to the lab is.
Grade Is Not the Same as Behavior
Even a perfectly representative assay only tells you what’s in the load, not how it will act. Two loads at the same headline purity can differ in ways the certificate doesn’t cover:
- Impurity mix. The trace elements that ride along matter. In dead-burned magnesia grades, for example, refractoriness is driven as much by the type of impurity as by the total MgO number on the sheet.
- Particle shape and size. Angular fragments pack differently than rounded ones. Fines behave differently than lumps in a screw feeder or a rotary kiln, even at identical chemistry.
- Mineral liberation. Whether the useful phase is locked inside host rock or freed from it changes reactivity, grindability, and yield, and rarely shows up on a routine assay.
- Moisture and surface condition. Two loads at spec moisture can still feed unevenly if one absorbed rain in transit and the other didn’t.
Closing the Gap Without Closing the Plant
No buyer is going to audit every load down to the grain. But the gap narrows when sampling is designed rather than inherited, when receiving specs include the properties the process actually cares about, and when a broker or sourcing partner is treated as part of the quality loop rather than a step upstream of it. A short conversation about who sampled what, where, and how, before the truck tips, tends to catch more surprises than a tighter tolerance on the same old certificate.
The plant floor isn’t being unreasonable when it says two identical loads behaved nothing alike. It’s reading a signal the paperwork wasn’t built to carry.
