Shingo distinguished internal setup, work that can be done only when the machine is stopped, from external setup, work that can be done while it runs. His conceptual stages follow from the distinction: observe a changeover in which the two are mixed, separate them, convert internal work to external, and streamline all operations (Shingo, 1985). Later practitioners extended the toolkit with waste-elimination routines for the setup crew (Sekine and Arai, 1992).

A group at the University of Bath examined the methodology critically over several years. Their central points are three. Changeover does not end when the new tool is fitted but when the line reaches full output at acceptable quality, so the run-up period belongs to the problem (McIntosh et al., 2000). Improvement can come from rearranging tasks, which is what the separation of internal and external work achieves, or from altering tasks so that they take less time, and Shingo's stages give too little weight to the second (McIntosh et al., 2007). And much of the second kind depends on the design of equipment and tooling, so organisation-led and design-led improvement should be weighed against each other from the start (Mileham et al., 1999; McIntosh et al., 2001). Design rules for changeover-capable equipment have been set out by Van Goubergen and Van Landeghem (2002).

Published applications show that the method must be fitted to the process. Tailored versions have been described for welding cells (Ferradás and Salonitis, 2013) and for automotive component production (Cakmakci, 2009), and Moxham and Greatbanks (2001) argue from a textile plant that teamwork, visual control and performance measurement are prerequisites without which the technique does not take hold.

A mould change on a horizontal machine involves more than swapping steel. The outgoing mould is drained or blown clear of coolant, its hoses, hot-runner cables and ejector coupling are disconnected, the clamps are released and it is lifted out by crane. The incoming mould is lowered between the tie bars, located on the fixed platen, clamped, and reconnected. If the material or colour changes, the barrel is purged and its temperature profile reset, and hygroscopic resins such as polyamide and polycarbonate must already have been dried. The mould is brought to temperature, process settings are entered, and shots are made until parts pass inspection (Rosato et al., 2000). In the vocabulary of total productive maintenance the whole interval counts as setup and adjustment loss, one of the six big losses that reduce equipment effectiveness (Nakajima, 1988).

The plant first agrees what is being measured. Following McIntosh et al. (2000), the proposed definition is the elapsed time from the last good part of the outgoing job to the first part of the incoming job that is produced at the planned cycle time and passes inspection. The definition deliberately includes heating, purging and trial shots. A baseline is taken from several ordinary changeovers on the chosen machine, recorded without special preparation, together with the material scrapped during each. A target is set, but it is expressed as a direction and a review date and not as a promise of single minutes.

One complete changeover is filmed with the setter's agreement, and the film is reviewed by the people who do the work. Each element is listed with its duration, who performed it, what tools and aids were used, and where the setter walked. Waiting is recorded as carefully as work: waiting for the crane, for a fitting, for the dryer, for the mould to heat, for quality approval. The output is an element sheet in which each line is marked as internal or external as currently performed.

The cheapest gains come from doing before the stop what does not need the machine. The incoming mould is inspected, fitted with its lifting eye and brought to the machine on a trolley. Hoses, clamps, tools and the setting sheet are laid out on a dedicated cart. Resin is in the dryer early enough to be ready. The crane is booked. After the restart, the outgoing mould is cleaned, protected and returned to its place while the machine runs. A checklist is the main instrument at this step, and a current-state map of the kind described by Rother and Shook (1999) helps to show planners how changeover time governs batch size.

Some tasks are internal only because of how the equipment is arranged. Mould heating is the leading case: a mould that is brought to temperature on a preheating station, or connected to a temperature controller before it is hung, removes a long wait from the stopped period. Hoses can be assembled into manifolds off line so that one connection replaces many. Process settings can be stored as recipes and loaded instead of typed. Table 1 lists the main elements of a mould change with the organisational and the design measure that applies to each.

What must be done at the stopped machine is then made quicker. Parallel working, with a second person on the non-operator side, removes much of the walking around the machine. Bolts of one size, slotted clamps and a single wrench reduce handling. Adjustment is treated as avoidable: if every mould in a family has the same back-plate dimensions and locating ring, the machine needs no change of mould height or ejector stroke between them. Here the choice between organisation and design becomes explicit. Quick-clamping systems and multi-couplers cost money per mould as well as per machine, and a small plant should cost them against the changeover frequency of the specific machine before buying (McIntosh et al., 2001).

In moulding the run-up is often the longest part of the changeover and the least visible, because the machine is cycling and appears to be working. Three measures are proposed. Jobs are sequenced so that consecutive jobs share material and move from lighter to darker colours, which shortens purging. The process settings from the last good run of each mould are stored and reloaded, so that the setter starts from a known point. And the criteria for releasing the job are agreed in advance with quality staff, with limit samples at the machine, so that approval does not wait for a person.

The improved sequence is written as standard work with photographs, the setters and their deputies are trained in it, and changeover time under the Step 1 definition is recorded for every change and displayed. Periodic observation checks that the standard is followed and collects ideas for the next round. The literature on prerequisites suggests that a changeover technique does not take hold without measurement and visual control of this kind (Moxham and Greatbanks, 2001).