A production line at steady state and the same line in its first shift are not the same system. In the first shift, nothing is tuned. Fixtures are being adjusted, programs are being proven, and operators are learning where the awkward features are. The output looks like the final product and behaves nothing like it.
This is why ramp-up deserves its own plan. A buyer who treats the first delivery as the first data point will draw the wrong conclusion from it, either that the supplier is poor or that the process is fine. Neither is knowable from a thousand units made under conditions that will never repeat.
Ramp-up is a learning curve with a cost attached, and the cost lands on whoever has committed to a date that ignores it. On welded assemblies such as stainless exhaust systems, that cost is concentrated in the first fixture.
The cost of learning is not mainly scrap, though scrap is the visible part. It is time: the time to adjust a fixture that is close but not right, the time to rewrite a program section that leaves chatter on one feature, the time to agree what a borderline measurement means.
That time is consumed by the most experienced people in the building, which means it is also time not spent on established work. A shop running several ramp-ups at once will feel this as a general slowdown, and the established jobs will slip even though nothing about them changed.
A planner who knows this will stagger introductions rather than launch them together. A buyer who knows it will ask how many other new parts are in ramp-up during the same period, and will treat the answer as a risk factor.
There is a stage between quotation and volume that many buyers never see: process proving. It answers whether the intended route can hold the drawing at the intended rate, and it produces the evidence that the first production run will be judged against.
Proving usually involves a small run under production conditions, with the real fixture, the real program, and the real inspection method. The output is measured, and any dimension that is marginal is either corrected or accepted with a documented reason.
A shop that skips proving and goes straight to volume is using the customer's order as its trial. That can work for simple parts. For anything with a tight feature, a cosmetic requirement, or a weld that has to pass inspection, it converts a controllable risk into a delivery risk.
When ramp-up runs late, the cause is more often tooling than machining. A fixture that was designed from the drawing without seeing the part will usually need adjustment on first contact, and the adjustment is iterative.
This is not a failure of skill. It is the normal cost of converting a two-dimensional specification into a three-dimensional holding arrangement. The parts that come off a new fixture are measured, the fixture is shimmed or re-clamped, and the cycle repeats until the measurements sit where they should.
Each iteration consumes a machine slot, which is why ramp-up competes with production for the same capacity. A shop that plans ramp-up outside the production schedule is planning fiction. The honest approach is to reserve machine time for learning, and to tell the buyer that the time is reserved.
The structure of a first delivery determines how much the buyer learns. A single large shipment at the end tells you whether the batch passed, and nothing about the process.
A staged release tells you much more. A small first article confirms the setup. A second, slightly larger release confirms that the process repeats. The balance follows once the measurements are stable. Each stage is a checkpoint, and each checkpoint is cheap compared with a rejected final batch.
Staging also protects the buyer's own commitments. If the first stage reveals a problem, there is still time to correct it before the downstream date. If the only stage is the last one, the correction happens after the deadline.
Ramp-up health can be observed from outside, without visiting the floor.
None of these requires disclosure of internal figures. Each is a behaviour, and behaviours are observable.
Buyers sometimes make ramp-up harder by asking for changes during it. A drawing revision in the middle of proving invalidates the fixture work and restarts the learning, and the cost appears as a delay that nobody planned.
The discipline is to freeze the specification for the duration of proving and release changes afterwards, as a separate exercise. If a change is genuinely necessary, it should be treated as a new introduction, with its own proving run, not as an amendment to the current one.
Similarly, a buyer should not compress the staged release to gain a few days. The compression does not remove the learning; it moves it into the production run, where it is more expensive and more visible.
Ramp-up risk is shared, and pretending otherwise is what makes it damaging. The factory carries the cost of learning; the buyer carries the cost of a late first delivery. Both are reduced by the same things: a stable specification, a proven route, a staged release, and a reserved slot for learning.
For parts where the process is unfamiliar, the first thousand units will always be the hard ones. A shop that says so is more trustworthy than one that promises a smooth start. On any new design, the first release should be treated as a proving run rather than a production run. The company background, the customisation process, other fabricated items under other parts and titanium components, and the dealership terms set out the rest.
