Two hours of changeover, the line stopped, the supervisor watching the clock. Most of that time gets lost because nobody separated, before the changeover started, what truly needs the machine stopped from what does not.
SMED, short for Single Minute Exchange of Die, is a method for cutting changeover or tooling time by separating the activities that can only happen with the machine stopped from the ones that can be prepared while the machine is still running. Shigeo Shingo developed it, and the stated goal is bringing changeover under ten minutes, which is where the single digit name comes from.
The split that changes everything
Those two categories are called internal setup and external setup. Internal is whatever requires the machine stopped: removing the old die, mounting the new one. External is everything else: locating the tools, staging the next die, checking that it is complete. In most plants I walk into, external setup happens with the machine already stopped, simply because nobody organized it ahead of time.
That is where the biggest and cheapest gain sits. Moving activities from internal to external is paid for with organization and no capital. It is the same principle I work in 5S, applied to the changeover cart instead of to the whole station.
What actually needs converting, and what just needs ordering
After separating comes converting: activities that felt internal out of habit but become external with a small design change. Here is an illustrative example: a line that takes 90 minutes to change over can lose 40 of those minutes just searching for the correct torque on each bolt. Standardizing that number and marking it on the tooling turns that search into external setup, with no new equipment purchased.
What stays internal gets simplified with quick clamping mechanisms, guides that only allow one position, and stops that remove the need for eyeballing an adjustment. This connects directly to poka yoke: a stop that only lets the die go in one way is not only faster, it also prevents mounting it backwards.
Why the minute gets lost afterward
The changeover gets designed once and measured well that first time, stopwatch and all. The problem shows up six months later, when the night shift no longer follows the documented sequence because nobody audits whether the order still matches what was designed.
An unaudited SMED looks a lot like an unaudited 5S. It degrades quietly, one shortcut at a time, until the changeover takes as long as it used to and nobody remembers why.
We covered how to split it, what to convert and why the gain fades over time. What is left is the part that actually decides the final number.
The number that actually matters
The single digit target is not always literal. What matters is measuring the current changeover with a real stopwatch, not with the number everyone repeats from memory, and measuring again after every round of improvement. Without that measurement, an SMED effort turns into a series of changes that feel faster and that nobody can prove actually are.
Beyond the time itself, I aim for repeatability: the changeover taking the same length whether a new operator runs it on Monday or the most experienced one runs it on Friday, because the sequence no longer depends on one person’s memory.
- What happens today with the machine stopped that could be prepared beforehand?
- Does the tooling have a single possible adjustment point, or does it depend on the operator’s eye?
- Was the last changeover measurement taken with a stopwatch, or from memory?
Start here
On your next changeover, stand next to it with a stopwatch and log every activity in two columns: what requires the machine stopped and what does not. That list alone, before changing a single thing, will already show you the first minute you can recover.
If you want to run a full SMED on your real changeover, with your team timing and redesigning the sequence, tell me how long your changeover takes today and which product triggers it most often.
What does SMED stand for?
Single Minute Exchange of Die: tooling changeover in a single digit number of minutes. It is a method for cutting changeover time by separating what requires the machine stopped from what can be prepared beforehand.
Does SMED always get changeover under ten minutes?
Not always literally, and that is not a mandatory target for every line. What the method does require is measuring with a stopwatch, separating internal from external, and converting what can be converted, round after round.
What is the difference between internal and external setup?
Internal is whatever can only be done with the machine stopped, such as mounting the new die. External is everything that can be prepared while the machine keeps running, such as staging tools or checking the next die is complete.