Below the splitting limit

Posted on: 5 November 2026

Heraeus, the family-owned German group based in Hanau just outside Frankfurt, is among the largest platinum refiners in the world. This week the Financial Times reported the findings of an independent investigation which concluded that between 2015 and the end of 2025 the company systematically reduced the results of assays on material belonging to its customers, almost always by less than 1 per cent. Around 450 customers worldwide were affected and some lost millions of euros. Heraeus has set aside €170m to compensate them, on top of the €460m already provisioned for customer metal retained in its plants, whilst German prosecutors are investigating sixteen current and former employees for fraud and embezzlement.

Most of the material in question comes from recycling, above all spent catalysts from industry and from cars, in which platinum and palladium are present in tiny proportions but at very high value. The customer delivers the material, the refiner samples it and assays it, then pays out in cash or returns in fine metal whatever the assay says was there. The entire commercial relationship therefore rests on a single number produced in a laboratory, namely the percentage of precious metal in the lot.

The obvious conclusion is that whoever measures the value of what he is buying will always have an incentive to err in his own favour. It is true but it explains rather little, because the customers were measuring too. Frank Stietz, the group's chief executive, has said that the deductions were originally introduced precisely to avoid discrepancies between the customer's assays and the refinery's own. To understand how a system with two independent measurements failed to notice anything for ten years, one has to look at how those two measurements are reconciled.

In the precious metals trade each party assays a sample of the same lot. If the two results diverge by more than a tolerance written into the contract, known in the trade as the splitting limit, a third sample goes to an independent umpire laboratory, and the cost of that analysis falls on whichever party strayed further from the umpire's result. If the difference stays within the tolerance, the two assays are simply averaged and the lot is settled. For platinum catalysts standard industry practice sets the splitting limit at 1 per cent relative. It is a sensible design, since it makes quarrelling over trifles expensive and penalises whoever overreaches. It does, however, contain an area into which nobody looks, and that area is precisely the space beneath the limit.

Put the two figures side by side. The deductions at Heraeus were almost always below 1 per cent, whilst the standard tolerance in the sector is 1 per cent. I have not seen the individual contracts and cannot claim that every deduction fell within the limit agreed in each of them. A gap of that size is nonetheless exactly the kind of difference that the umpire mechanism is built not to catch, given that below the limit the procedure itself prescribes averaging and moving on. A refiner who lowers his own assay by 1 per cent thus pockets half of it, or all of it if the customer trusts him and does not bother with an assay of his own. A fraud has no need to circumvent the control if it is sized to the control's tolerance.

The industry knew about this risk and had even put it in writing. For years consultants and the technical press have warned that a refiner can offer apparently generous commercial terms and recover the difference by trimming the platinum content by a fraction of a point, and they have flagged less scrupulous refiners trying to push the splitting limit to 5 or even 10 per cent. The standard advice to customers is to divide their material into many small lots rather than a few large ones, so that each is measured and reconciled separately. The weakness had a name, but for a decade nobody went looking for it in the house of one of the most respected names in the business.

A continental eye that has spent decades inside organisations recognises the pattern wherever an accepted margin exists, from the expense claims that stop just short of the amount requiring a receipt to the invoices kept one pound below the threshold for a second signature. Every verification system has to tolerate some error or it seizes up, and whoever knows that margin can take everything that fits beneath it. Half a point on a single lot is nothing, yet across thousands of lots and hundreds of customers over ten years the total runs to nine figures.

There is a further detail which strikes me as more instructive than the fraud itself. It was not the customers who uncovered the problem. It was Heraeus, after finding more metal in its own inventories than its records could account for, and from there the crisis widened until the company acknowledged that the surplus belonged largely to its customers. Numbers can be adjusted for years, whereas retained platinum sits in a warehouse until the materials balance stops adding up. In the end the control that worked was not the one written into the contracts but the physical one, which is also the oldest and the least sophisticated.

British readers have an older answer to this problem than most. Hallmarking in England goes back to a statute of Edward I in 1300, and under the Hallmarking Act 1973 precious metal articles sold as such must be tested and marked by one of the independent assay offices in London, Birmingham, Sheffield and Edinburgh. The principle is structural rather than contractual, since the party who certifies the fineness is neither the maker nor the buyer. Switzerland, which refines a large share of the world's gold, took a different route with its 1933 law on precious metals control, entrusting the assay of melt products intended for resale to sworn assayers who hold a federal diploma and a personal licence that can be withdrawn. One system places the guarantee in an institution with nothing to gain from the result, the other in an individual with something to lose.

Neither should be mistaken for a complete answer. Hallmarking concerns finished articles on their way to the consumer, not the settlement terms on which a refiner pays for scrap coming through its gates. The Swiss sworn assayer often works inside the refinery itself, which can hold both the smelting licence and the authorisation to assay, so his signature raises the cost of lying without removing the incentive, and the Swiss Federal Audit Office found that until early 2019 supervision of these operators was limited to brief administrative checks. The recycling settlement, which is exactly where Heraeus fell, remains largely a matter between two private parties and the tolerance they agree.

The lesson to draw from this affair is that anyone entrusting valuable material to a third party should ask less whether his own controls are rigorous and more what everything that slips beneath them is worth, added up over the years.