How a product creates value
The ship date is fixed early. Cost and risk are not.
If the ship date is fixed, what does engineering still decide?
On many hardware projects, the mass-production date is set early. Engineering rarely moves that date. What it shapes is what the product ships with: its cost, its risk, and what the next project can learn.
Three mechanisms shape that. None of them appears on a test report.
Cost is committed before it is spent
Material, process, tolerance, part count: each is chosen in design, and each sets a cost the factory pays on every unit. Much of a product's cost is decided before most of it is spent.
The figure usually quoted is that design commits about 70% of a product's cost. The exact number is disputed, and so is how much design alone decides. Still, many cost choices are locked in early.
The best time to reduce cost is early in design.
Each build stage makes changes more expensive
NPI moves in builds. EVT tests the design, DVT tests it against the specification, PVT tests whether the line can build it. Each build commits more: tooling is cut, parts are ordered, lines are set up.
A change that is a drawing edit before EVT can mean new tooling after DVT, and a recall after ship. Quality engineers call it the rule of ten: a fix costs roughly ten times more at each stage it survives. The exact multiplier varies by product and stage, but the direction holds.
The date rarely moves to make room for a late problem, and when it does, that has a price too. The problem is either fixed at a higher cost, or it ships as risk.
The later the change, the more expensive the decision.
Small yield losses cost more at high volumes
At a million units, 99% and 99.9% yield mean 10,000 failures versus 1,000.
Yields multiply. Ten process steps at 99% first-pass yield each give about 90% end to end. Twenty give 82%. A step that looks fine on its own still lowers the total.
Confidence has to survive scale. So does value. A defect rate is a unit cost, multiplied by volume.
At a million units, every 0.1 percentage point of yield is 1,000 units.
The question I bring
I am early in my career, and the ship date is not mine to set. What is mine is one question I bring to every result and every review: if this is wrong, where will we find out, and what will it cost by then?
Every role I've had answers part of it. Learning to answer all of it is the engineer I'm working to become.