Most quality problems do not start as disasters. They start as a small deviation that everyone involved believed was close enough. A dimension a hair outside tolerance. A material that tests fine on average. A step skipped because the last thousand parts were perfect. None of it looks like a problem at the moment it happens. That is exactly what makes it dangerous.
In 1628 the Swedish warship Vasa sailed on her maiden voyage, heeled over in a light gust, and sank in about twenty minutes, barely a mile from shore. What stays with us is not that the ship was flawed. It is that they already knew. A stability test had been run before she sailed: thirty men ran back and forth across the deck, and the ship rocked so violently the test was stopped for fear it would capsize at the dock. The data was there. They sailed anyway. The failure was not the flaw. It was ignoring the signal that the flaw was sending.
That is the real subject of quality assurance. Not catching the obvious defect, but respecting the small signal before it becomes an expensive one. After years of supplying one critical component to manufacturers on five continents, here is what we have learned about where quality is actually won or lost.
Almost right is not a little bit wrong. It is a different category A part that is two percent out of spec does not perform two percent worse. On something that moves, it can work perfectly for a while and then fail all at once, or worse, fail intermittently in a way no one can reproduce. The gap between "in spec" and "almost in spec" is not small. It is the gap between a part you can trust and a part you have to worry about.
You cannot inspect quality in at the end. Final inspection is the last net, not the first. If quality is not built into the process that makes the part, no amount of checking at the end will put it there. Inspection at the end catches symptoms. Capability in the process is what removes the cause. A supplier who leans entirely on final inspection is telling you they are managing defects, not preventing them.
The dangerous defect is the one that passes. The failures that hurt most are not the parts that get caught. They are the marginal ones that clear inspection and fail later, in an assembly or out in the field, where they are almost impossible to trace back. The goal is not a low defect rate that you catch. It is a process capable enough that the marginal part never ships in the first place.
A quality miss is quietest at the source and loudest in the field. The same deviation costs almost nothing to catch at the machine, something to catch at receiving, a great deal to catch in a finished assembly, and a small fortune to catch in a customer's hands. Nothing about the defect changes as it travels. Only the price of finding it does. Traceability and control at the source are not paperwork. They are the difference between a scrapped part and a recall.
None of this shows up on a certificate of conformance. It shows up in whether the same supplier delivers the same part, correct, run after run, on the ordinary Tuesday when no one is watching. That is the standard we hold ourselves to on one very specific component, for manufacturers around the world. We make something that lives inside a much larger machine, on things that move, and when it does its job, no one thinks about it. That is the entire point of it.
The question worth asking any supplier is not whether they can hit the spec once. It is whether their process is built so that "almost right" never leaves the building. Those are not the same capability, and the difference is every failure you never have to explain.