The price moved. The change notice didn't.

This week a wheat-flour price surge turned roti into a national crisis, and most of the commentary filed it under inflation. I file it under supplier quality, because I have lived the mechanism. Millers facing a cost spike blend cheaper grain. Bakers stretch the flour with whatever still passes for flour. Every certificate in the chain still describes the original recipe. Every incoming check still passes. And the flatbread leaving the tawa is no longer the product anyone qualified. Nobody was malicious. Nobody even lied. Each link in the chain defended its margin in the one place its customer cannot see: the bill of materials. I have stood at enough receiving docks to know how this film ends. Inflation headlines are back, and so is this failure mode – the same week, a fuel-quality investigation widened its probe upstream. Same pattern, different molecule.

The substitution nobody files

When an input cost doubles, a supplier has three moves: raise the price and risk the contract, absorb the loss, or find an "equivalent". Only the third keeps everyone quiet. The new mill, the new compound, the new ingot supplier arrives with a certificate matching the old one, because grades are bands, not fingerprints – two coils inside the same EN designation can behave like different metals on a press. Incoming inspection passes, because inspection tests to the drawing, and the drawing never mentioned the rolling mill. The change is real; the paperwork is silent. The change-control clauses of IATF 16949 sleep straight through it, because the trigger list – drawing revision, tooling change, process transfer, plant relocation – was written by engineers for engineering departments. VDA 6.3 auditors ask their change-management questions and get answers that are technically true. In most quality systems I have audited, purchasing and quality share a car park and nothing else.

Suppliers do not hide changes out of malice. They hide them out of arithmetic.

The PFMEA describes a material you no longer buy

Document drift is quiet. The shop floor is not. Forming behaviour moves first: burr height, edge cracking, springback. Heat-treat response shifts a few points on the hardness curve. Torque curves on lock barrels start wandering months before anyone files an 8D. APQP froze the inputs at design lock; a purchasing decision unfreezes them and tells no one. By the time the failure reaches a customer's dock, the PFMEA in your QMS describes a part that no longer exists – its occurrence ratings anchored to a material you stopped buying two price cycles ago.

I learned this at Witte Automotive, the hard way. The certificates were clean; the press-force signatures were not. A QRQC team chased a drift in forming data back through six weeks of coils and landed on the truth: the supplier's sourcing department had moved to a new rolling mill. Same grade designation, metallurgically a different animal. Nothing in the paperwork had ever been wrong. Everything in the process data had been shouting. The A3 that followed rebuilt incoming control around signatures rather than certificates – press-force windows, torque bands, hardness scatter – and that QRQC-and-A3 discipline delivered the substantial failure-cost reduction I still point to. The certificates warned us exactly once: never.

Price belongs on the quality dashboard

At SNOP I built a greenfield quality organisation for a plant of more than 900 people, which has one advantage nobody mentions: you write the rules before the habits exist. We treated incoming material as a live process, not a verified fact – surveillance sampling keyed to risk, escalation rehearsed like a fire drill, no document accepted as a substitute for data. That discipline produced a 70% reduction in defect costs and a quarter with zero critical customer escalations. Containment beats paper conformity every time the two are tried against each other.

I have also sat on the wrong side of this failure, de-escalating a customer after an "equivalent" grade reached their line. That conversation is never about metallurgy. It is about trust, and it is won with data, speed and candour – hours, not weeks – or not at all. So wire the commercial triggers into the quality system before the escape, not after:

• PCN clauses with teeth. Any raw-material source change, or any indexed price move beyond an agreed threshold, obliges notification within days and re-PPAP on affected part numbers, with contractual consequences for silence.
• Surveillance sampling keyed to price. When the commodity index spikes, tighten incoming sampling for that material family before the first suspect truck arrives.
• Dual sources that actually run. An approved second source that never sees volume is a rumour. Move real parts through it deliberately and keep its requalification current.

The Monday move is deliberately unglamorous: wire the commodity dashboard into the quality alert system. When a designated material's indexed price moves more than 15% in a quarter, a change-management ticket opens itself, surveillance sampling tightens, and re-PPAP is requested for every affected part number – before the next truck ships. I spend part of my time designing autonomous multi-agent monitoring systems, and this end of the problem is one integration and a workflow, not a transformation programme. The principle is simple: commercial changes must clear the same requalification gates as engineering changes, because the customer's press, oven and torque wrench cannot tell a drawing revision from a purchasing decision. A requalification run costs thousands of euros. A field escape costs a customer. In the roti case, it cost a country its breakfast.