One-off production: making a thing exactly once

Verification happens while the part is still on the machine

Every dimension, weld and finish on a single unit is produced for the first and last time, so inspection cannot be a sampling exercise. In-process checks, dimensional verification before a feature becomes inaccessible, and a written record built as work proceeds are the only realistic controls. Rework here is usually repair rather than replacement, which brings its own concession and approval paperwork. The commitment the shop makes is to attention rather than to throughput: a job reaching final assembly with an uncaught error may have to be cut apart to fix.

Estimating in the absence of history

Quoting is the hardest discipline in the model, because reference data thin out exactly where the money is. Estimators build from elemental times, comparable features on earlier jobs and an explicit allowance for the unknown, then learn the truth once. Systematic under-quoting shows up as a busy shop with no profit, which is why comparing estimated against actual hours after despatch, and feeding that back into the estimating basis, matters more here than anywhere else. The work that suits this shape is urgent, unique, or too awkward for a volume supplier to want.

General machines, specific people

Capital goes into versatile equipment — manual and computer-controlled machining, welding across several materials, forming, access to heat treatment, measurement capability — chosen for range rather than cycle time. None of it is worth much without trades able to use it, and that is the ceiling on growth: the shop expands only as fast as it can recruit or train people who can work from a drawing without a proven method sheet. Extra machines simply queue behind the same individuals, which is why apprenticeships and documented technique are commercial matters, not merely cultural ones.

Buying material for a single piece

Material is bought to the job, so there is no stock to draw on and no volume to negotiate with. Mills and stockholders sell in their own increments, meaning one plate or a length of bar arrives with an unusable remainder the job must carry. Certification tightens it further: a traceable heat with test documents may only exist in a quantity far larger than the piece needs. Suppliers willing to cut, certify and deliver small amounts quickly become the shop's real dependency, and their price premium is properly a cost of the model rather than a purchasing failure.

Where single-piece work goes wrong

The characteristic failures are a drawing misread at the start and discovered at assembly, an unpriced scope conversation with the customer's engineer, and a quotation that never covered a setup nobody gets to repeat. Revision control matters disproportionately, since working from a superseded drawing wastes the whole job rather than one batch. Systems should stay simple but real: job costing capturing hours against operations, a document register showing which revision reached the floor, and a photographic and dimensional record retained after despatch, because a query arriving later cannot be settled by re-measuring the part.

Frequently asked questions

How do we quote work we have never done before?

Break it into features rather than pricing it whole. Setups, machining volumes, weld lengths, handling operations and inspection steps can each be estimated from experience even when the assembly is unfamiliar, and building the figure that way makes the assumptions visible. Add a stated allowance for the unknown and tell the customer what the quotation excludes. Then compare outturn against estimate on every job, so the basis improves instead of repeating the same optimism.

Is it worth holding any stock in a shop that makes singles?

Only for genuinely common consumables and standard sections appearing in most jobs — fasteners, welding consumables, abrasives, a handful of frequently used bar and plate sizes. Beyond that, buying speculatively ties cash up in items that will not match the next drawing. What is worth maintaining instead is supplier relationships and a certified offcut register, so a piece already on site with valid documentation gets found before another one is ordered.

How much documentation does a single unit really need?

Enough to answer a question after the unit has left the building. That normally means the released drawing revision, material certificates traced to the piece, records of any concession or repair, inspection results against the specified features, and evidence of any functional test performed. Regulated or safety-related work will demand more. Assembling it during the build costs little; reconstructing it afterwards, once the welder has moved on and the offcut is gone, is often impossible.