The situation

The molding machine stops at 06:00 for a job change. At 06:03 the setter walks to the warehouse for crane chains. At 06:12 he looks for the grey resin. At 06:20 the mold is on the machine, and then it waits for the mold to heat. The supervisor says “we need to be faster at changeovers.” The setter has not stopped moving all morning. Faster is not the problem.

What SMED is for

A changeover is the time from the last good part of one product to the first good part of the next. The machine makes nothing during that time.

SMED means Single-Minute Exchange of Die. It is a method for shortening changeovers by changing the order of the work, not the speed of the people. Its core idea is a split:

  • Internal work can only be done while the machine is stopped. Example: unclamping the mold.
  • External work can be done while the machine is still running the old job. Example: fetching tools and material, pre-heating the new mold.

Most changeovers do much of their external work after the machine has stopped. That is where the time goes. Moving it to before the stop is usually the biggest, cheapest gain.

“Single minute” means single digits: under ten minutes. It does not promise one minute for every machine.

A short history

Shigeo Shingo, a Japanese industrial engineer, developed SMED between 1950 and 1969. The sources we use say he first saw the internal and external distinction in 1950, during a study of a large press at Toyo Kogyo (Mazda). They report that in 1969, at a Toyota stamping plant, an 800-ton press setup fell from about four hours to under nine minutes. We have not checked these figures against the original, so read them as reported. Shingo described the method in A Revolution in Manufacturing: The SMED System (Productivity Press, 1985).

Where it applies, and where it does not

It fits:

  • high-mix, low-volume production, where long changeovers push people to run big batches (stamping, molding, machining, roll forming);
  • packaging and bottling lines with frequent product or label changes;
  • processes that need purging or warm-up between jobs, such as paint and ovens.

It fits badly:

  • Single-product lines that never change over.
  • Manual assembly where “changeover” means a different parts bin. Use 5S and layout instead.
  • Software deployment. That is a different problem.

How to do it

Our example is a fictional plant, Krosno ThermoPlastics. Dariusz is the shift leader. Karolina is the process setter. Injection molder IM-04 changes from a black cup-holder bezel to a grey door pocket insert.

  1. Observe and write it down. Film one complete changeover, last good part to first good part. List every step and how long it took. Karolina’s list totals 68 minutes: 18 minutes finding crane chains, grey resin and clamp bolts; 16 minutes waiting for the new mold to reach 85°C; 24 minutes undoing and tightening 16 threaded studs; 10 minutes adjusting pressure and running 18 test parts to clear flash.
  2. Separate internal from external. For each step ask: does the machine really need to be stopped for this? Fetching chains and resin: no. Heating the mold: no. Removing studs: yes. Adjusting: yes, as it is done today.
  3. Convert internal to external. Stage the resin, chains and tool cart 20 minutes before the old run ends. Put a portable water heater on the new mold so it is at 85°C before the machine stops. That removes 18 and 16 minutes from the stoppage.
  4. Shorten what is left inside. Replace the 16 threaded studs with pneumatic quarter-turn wedge clamps. Clamping drops from 24 to 4 minutes. Add calibrated mechanical stops and a stored program recipe, so there is no dial hunting. One test shot takes about 2 minutes.
  5. Organise the outside work. Use a changeover cart, shadow boards and marked floor squares, so the next job’s items are always in the same place.
  6. Write it down and test it. Make a one-page visual procedure with who does what. Run it, time it, adjust it.

The new total is 4 minutes clamping plus 2 minutes for the test shot: 6 minutes. Down from 68, that is 62 minutes saved, or 91%. (Our source research quotes 8 minutes for this example; our hand count of the listed steps gives 6. If a real changeover has other steps, such as a first-part inspection, add them.)

Two people working at the same time on the remaining internal steps can save more. The goal is a choreography, not a sprint.

Three ways it goes wrong

  1. The pit-crew fallacy. Management asks people to run faster. They strip threads, drop tooling and pinch fingers. Safety gets worse, and the time hardly improves. The fix is fewer steps, not faster steps.
  2. No pre-staging. The machine stops, and then someone looks for the crane hook, the forklift or the torque wrench. A large share of downtime is walking and searching. A simple checklist before the stop fixes it.
  3. Trial and error at the machine. Test cuts, calipers and dial tweaking can eat a large part of a changeover when settings are not fixed. Use fixed stops, digital readouts and stored recipes.

Our recommendations

  • Film it. People remember changeovers as shorter than they are.
  • Do the observation with the setter, not about the setter. They know where the time goes.
  • Start with external conversion. It costs the least.
  • Re-time after each change to confirm it worked. This is the PDCA Check step.
  • Count the change in OEE terms: setup is a large part of availability loss.

Try it

The changeover tool will let you log steps, mark them internal or external and see the before-and-after timeline, plus a printable card for the machine. It is coming soon: changeover tool.

From the book

Changeover (SMED) is Chapter 14. Its motto: do not run faster, prepare before the machine stops. See the book page.