Improvement in a jobbing shop: what repeats when the products do not

The repeating unit is the operation, not the product

A machine shop making several hundred different parts a month performs the same small set of activities on all of them. Setting a machine, proving a program, checking a first-off, changing tooling, cleaning and deburring, and preparing paperwork are common to every job, and each has an established best method currently carried in the heads of individuals. Documenting and improving those applies across the whole order book without needing any product to repeat. This is where standard work has meaning in a low-volume environment, and in most such shops it is entirely absent.

Standard work for setting, not for cycles

The repetitive-plant version, with a fixed sequence, a fixed cycle time and a fixed quantity of work in process, does not transfer. What does transfer is a defined method for each recurring activity: how a first-off is presented and checked, where tooling is staged before a setup begins, what the setter does while the machine is running, how a job is handed to the next operation, and what the operator records. Those are the activities that consume the shift in a low-volume shop, and standardising them is felt immediately by everybody working there.

Which methods travel and which do not

Pull signals tied to a part number are meaningless for one-off components, and levelling has nothing to level. Setup reduction, mistake-proofing, workplace organisation, visible job status, problem-solving discipline and flow through the office all transfer directly and frequently pay better than in volume plants, because the starting point is worse. Card-based replenishment does apply to the consumables and standard materials every job draws on, such as fasteners, bar stock, abrasives and gases, and that is usually the least contentious place to begin, since nobody argues that stores should run out.

Queue discipline replaces the pull signal

With unique parts, the control question becomes how much work is on the shop floor at once and in what order it is worked. Releasing every order as soon as it is entered fills the shop with jobs that then wait, obscuring priority and inviting expediting. Limiting released work to what the shop can start, and enforcing a visible sequencing rule supervisors do not quietly override, achieves most of what a pull system achieves in repetitive production. The hard part is political, because the released backlog is where informal favours to customers currently live.

The biggest losses sit before any metal is cut

In engineered and one-off work the elapsed time is dominated by quoting, drawing interpretation, technical queries, material procurement and waiting for a decision. A job can sit for weeks awaiting an answer that takes minutes to give. Shops measuring only machine utilisation never see this, because the machines were busy on other work throughout. Following a sample of jobs from enquiry to despatch, and recording where each one waited, redirects improvement effort towards the office, which is uncomfortable since that is where the owners and engineers usually sit.

Frequently asked questions

Can card-based replenishment work when every part number is used once?

Not for the parts themselves. It works well for standard materials and consumables that every job draws on, and for tooling and inserts, where a simple two-bin arrangement removes stockouts and the constant small interruptions they cause. For the jobs, the equivalent control is limiting how much work is released to the floor and sequencing it visibly, which delivers the same effect of stopping the shop from being flooded with work it cannot start.

Is standard work realistic when every job is different?

Yes, at the level of the operation rather than the product. Write down how a setup is prepared, how a first-off is verified, how tooling is staged and how a job is handed over, and you have standardised most of what actually consumes time. The test is whether two setters approach the same machine the same way. Where they do not, that difference is unmanaged variation and it costs time on every job regardless of what the drawing shows.

Where should a small high-variety shop start?

With setup reduction on the busiest machines and with the flow of information through the office, in that order or together. Both produce visible results without capital, both apply to every job, and both are unarguably wasteful in their current state. Workplace organisation is a reasonable third, provided it is aimed at finding tooling and materials quickly rather than at appearance. Avoid launching a plant-wide programme, since small shops have no spare management attention to run one.