Lean tools

Single Minute Exchange of Die

Q: What does SMED mean exactly?

SMED is the acronym of Single Minute Exchange of Die, literally "tool change in a single-digit number of minutes" — so in under ten minutes. The method was formalised by the Japanese engineer Shigeo Shingo from his work at Toyota. It is also referred to as quick changeover, or rapid tool change.

In French: SMED (changement rapide d’outil)

SMED cuts changeover time by separating what can be prepared while the machine runs from what requires it to stop, then converting the second into the first.

What it is

SMED stands for Single Minute Exchange of Die: literally, a tool change in a single-digit number of minutes, so under ten minutes. The method was formalised by Shigeo Shingo from his work at Toyota. The stated ambition is not a slogan: it sets an order of magnitude, not an obligation.

What is at stake goes beyond the time saved. As long as a changeover costs three hours, you produce in large batches to amortise it — so you carry stock, lead time and the risk of discovering a defect too late. Dividing the changeover time allows smaller batches. SMED is therefore first a lever on flow, and incidentally a lever on productivity.

The scope to measure is precise: from the last good part of the previous run to the first good part of the next, at nominal rate. Everything is inside — stop, dismantling, fitting, adjustments, quality validation, ramp-up.

The four stages of the method

Observe and separate — film a real changeover, then classify each operation as internal (machine stopped) or external (doable while the machine runs). That separation alone often produces a substantial reduction, with no investment.

Convert — turn internal into external: prepare and preheat the tooling, position the components, adjust off the machine.

Streamline — replace screws with clamps or quarter-turn fasteners, standardise heights and interfaces, prepare settings on a jig.

Eliminate adjustments — adjustment by successive trials is the biggest consumer of time. Replacing it with stops, marks or coded positions is what tips a changeover under ten minutes.

What it is for, and how to recognise the need

You launch an SMED project for a reason that is almost never "saving machine time". The real reason is that you want to produce shorter runs more often, because the customer orders in small quantities and the stock absorbing that gap is expensive — in cash, in floor space, and in defects discovered too late.

So the symptoms are to be looked for somewhere other than on the machine. A schedule frozen six weeks ahead because "we cannot change over more often". A warehouse sized to hold out between two runs of the same reference. A complaint about a defect that appeared three weeks earlier and was reproduced on four thousand parts. Three situations, one problem.

The payback is calculated on that basis, not on the minutes saved. A changeover divided by three makes it possible, at unchanged total changeover time, to triple the number of changeovers — hence to divide the batch size and the average work in progress in the same proportion. It is that conversion that must be presented to obtain the means for the project; the "minutes saved" line convinces only those who already agreed.

One print changeover, minute by minute

Label printing line, reference changeover filmed end to end. Stopwatch started on the last good label of the previous run, stopped on the first good label of the next at nominal rate: 96 minutes. The film gives five blocks — searching for tooling and the job file 14 min, dismantling 11 min, fitting 19 min, adjustment by successive trials 34 min, ramp-up and quality validation 18 min.

Separate. Searching for the tooling and the job file, 14 minutes, can be done entirely while the machine runs: a trolley only has to be prepared during the previous run. The changeover drops to 82 minutes, and it is the cheapest gain of the project.

Convert. Fitting was done piece by piece on the stopped machine. Pre-assembled on an off-line fixture, it falls from 19 to 7 minutes of machine downtime. We are at 70 minutes.

Streamline. The screws of the dismantling give way to quarter-turn clamps and the interface heights are standardised: 11 minutes become 5. We are at 64 minutes.

Eliminate adjustments. This is the 34-minute block, and it is where the project is decided. Mechanical stops and engraved marks replace the successive trials: 9 minutes. The ramp-up, which benefits from a setting that is right first time, drops from 18 to 11 minutes. Total: 0 + 5 + 7 + 9 + 11 = 32 minutes.

From 96 to 32 minutes, a factor of three, with no new machine. Yet the figure to remember is not that one: with three changeovers a week, the line was losing 288 minutes weekly; at the same budget of lost time, it can now do nine. It is that sentence that unlocked the move to short runs, not the 64 minutes saved.

One detail of the recording deserves a note: the 34 minutes of adjustment appeared on no report. They were counted as "production" because the machine was running — it was producing set-up labels, which were scrapped. Without the film, that block did not exist.

The common mistakes

Timing the changeover from the machine stop, and stopping the clock at restart. That leaves out the ramp-up and the set-up parts, i.e. often a third of the real time. The figure obtained is flattering and unusable: it corresponds to nothing the line actually lives.

Skipping the filmed observation and working on the sequence the team recounts in a meeting room. The account is always cleaner than reality: it leaves out the trips to the store, the tool being looked for, the colleague being waited for. Those are not omissions in bad faith, they are motions that have become invisible through repetition — and they often weigh more than the operation that was going to be optimised.

Reducing the changeover time and leaving the batch size unchanged. The project has then cost money to produce available machine time that nobody needed, and the stock has not moved by a single carton. An SMED ends in the production plan, not on the machine: if the batch sizes are not revised, the gain stays theoretical.

On the shop floor — Plastics processing

Simply preparing the next mould on a trolley while the press was producing — converting internal into external, without buying anything — took nearly forty minutes off the changeover. The following gains cost money; that one cost a reorganisation.

The other tools illustrated by this trade: plastics and chemicals.

Frequently asked questions

What does SMED mean exactly?

SMED is the acronym of Single Minute Exchange of Die, literally "tool change in a single-digit number of minutes" — so in under ten minutes. The method was formalised by the Japanese engineer Shigeo Shingo from his work at Toyota. It is also referred to as quick changeover, or rapid tool change.

What is the difference between internal and external operations?

An internal operation requires the machine to be stopped: dismantling the previous tool, fitting the next. An external operation can be done while the machine is still producing: preparing the tooling, pulling the job file, preheating a mould. The whole point of the method lies in that boundary — first stop doing at a standstill what could be done while running, then move the boundary itself.

Do you really have to get under ten minutes?

No. The ten minutes set an ambition and give the method its name; they are neither a regulatory threshold nor a success criterion. On a large press, going from six hours to ninety minutes transforms the schedule far more than going from twelve to nine minutes on a small piece of equipment. The right target is the one that allows the batch size the customer needs.