Every scrap purchase is settled in one unit and consumed in another. Purchasing closes on a landed price per ton of a named grade. The furnace consumes recoverable iron units, whatever alloying elements arrive with them, a residual load it cannot refine away, a yield loss it absorbs, and energy and time consequences it never got to negotiate. Those are two different numbers, and only the second determines the cost of a liquid tonne at the chemistry the order requires.
That gap is where the standing argument between purchasing and the melt shop lives. One is measured against a price it can defend line by line, the other against yield, power-on time and a residual cap. Both are honest; the unit of account was never translated.
What metallurgical value is made of
Metallurgical value, or value in use in the older procurement vocabulary, is what one charged tonne is worth to your shop, given your grade slate, your caps and your bottleneck. It decomposes into separable terms:
- Recoverable iron units. Not billed weight, but the metallic fraction that survives to the bath, valued at what an iron unit is worth to you, which your alternative source of iron sets rather than any published scrap price.
- Recoverable alloying elements, valued at the ferroalloy they displace. Nickel and molybdenum arriving in scrap are essentially not oxidised out of an EAF bath: they report to the steel and displace a purchase of ferronickel or ferromolybdenum. Chromium partly does not survive: under oxidising conditions a meaningful share reports to the slag, so it is worth less than its assay implies. The correct price for an element is the delivered cost of the alloy addition it replaces, adjusted for the recovery you actually get, and only while you are still short of aim. Nickel you did not need is not a credit; in a low-residual grade it is a problem.
- Residual load, valued as the dilution it forces. Copper and tin are the hard constraint. Neither can be oxidised out of the bath; both accumulate, and copper in particular drives hot shortness at the surface during hot working, which is why flat and drawing grades carry caps that long products do not. A copper-rich load therefore has a cost that is not on its invoice: the clean dilution material (prime scrap, busheling, DRI or HBI, pig iron) you now have to buy to stay under the cap. Past a certain point it also costs you optionality, because the grades you can still make from the pile narrow.
- Yield loss, because a charged tonne is not a liquid tonne. Oxidation of the charge, free iron oxide riding in on already-rusted material, non-metallics, dirt, fines and entrained moisture all separate billed weight from liquid weight. Yield is a property of the specific stream and supplier, not of the grade name on the ticket.
- Density, size and moisture, valued as energy and minutes. Low bulk density means more buckets per heat, which means more lid-open time, more radiation loss and a longer tap-to-tap. Oversized or awkward pieces disturb arc stability and delay scrap collapse. Moisture consumes energy to vaporise, and carries its own safety exposure. If the furnace is your bottleneck, minutes convert into forgone contribution; if it is not, they cost electricity and electrodes and little else. The same load is worth different amounts to the same plant in those two states.
Two loads at the same price are not the same purchase
Take two offers at an identical landed price per ton. One is a well-prepared, low-residual bundle: high metallic yield, poor density. The other is shredded material that handles well but carries more non-metallics and a copper level near the top of what your flat-rolled slate tolerates.
On the invoice they are interchangeable. On the value-in-use ledger they diverge on four terms at once: the number of liquid tonnes each produces, the ferroalloy credit each carries, the dilution each one obligates, and the minutes each costs. Those terms do not move in the same direction, and there is no rule that says the cheaper ton is the cheaper tonne of steel. Sometimes the right call is to pay a premium a buyer's scorecard will flag; sometimes it is to walk away from an apparent bargain because dilution capacity is already committed. Both are invisible when the only common unit is price.
And the divergence is not a fixed property of the two grades: it depends on what you are rolling next month, how much clean iron you already hold under contract, and whether the furnace is starved or saturated. Value in use is a number about your plant, not about the material.
What changes when the shop buys on value
The mechanical change is small: bids get compared on a common denominator. The consequences are not.
Purchasing and the melt shop stop arbitrating between two scorecards: the purchase decision and the charge-mix decision become one problem rather than two sequential handoffs. Dilution acquires an explicit price, which means low-residual iron (DRI, HBI, pig iron, prime industrial scrap) becomes a routine comparable rather than an emergency purchase made when a heat is already in trouble. The ferroalloy budget and the scrap budget stop being independent, because a nickel-bearing load is partly a ferroalloy purchase and should be approved as one.
Supplier evaluation shifts hardest. Consistency of delivered chemistry becomes a priced attribute, not a soft one: if a supplier's copper varies widely load to load, you have to charge as though every load is near the top of that range, so the variance itself has a cost even when the average is fine. That is a different conversation from a price negotiation, and one in which a yard that controls its inbound streams can be paid for doing so.
What makes this genuinely hard
It would be dishonest to present this as a solved measurement problem.
Chemistry is learned late. The bath sample describes a heat, not a load: several loads went into it, and the material is gone by the time the lab answers. Attribution back to individual purchases is retrospective, pooled and noisy, and it takes many heats before a supplier's true distribution separates from it.
Grade names describe form, not composition. Two loads of the same nominal grade from different yards can sit far apart on residuals and yield, and the spread between suppliers is often wider than the spread between grades. Any credible model is fitted per stream, which takes the discipline of recording which stream went where.
Residual caps are a portfolio constraint, not a per-heat one. Dilution capacity is finite and shared across the order book, so the true cost of accepting a copper-rich load includes what it does to heats you have not scheduled yet. That coupling is exactly what makes the intuition of an experienced buyer hard to beat in a single decision, and hard to sustain across a month of them.
Prices on both sides move on different clocks, too: scrap indices and ferroalloy markets are quoted at different frequencies, on different regional bases. A value-in-use figure is a decision aid built partly on stale inputs, and it should carry that uncertainty visibly rather than arrive as one confident number.
Where this leaves the buyer
None of this argues for abandoning $/ton: it is the unit the market trades in. It argues for not mistaking it for the thing you care about. The price is what you pay; the metallurgical value is what you bought.
Making that translation by hand for one load is straightforward arithmetic. Making it for every open offer, against a live residual cap, a live grade slate and a moving ferroalloy market, is the part that needs tooling. That is what our OptiScrap module does: it compares procurement options on value-in-use cost and recommends a mix rather than the cheapest ton.