Lean teams use SMED because long changeovers force big batches. Faster changeovers allow smaller lots, more frequent production of every product and less inventory.
Where it comes from
Shigeo Shingo developed SMED through work with Toyota and other Japanese manufacturers. It became one of the methods that made small-lot, just-in-time production practical.
At a glance
Changeover before and after SMED
Swipe sideways to see the whole diagram.
How to apply SMED
- 1
Observe and record the changeover
Video a full changeover and list every task with its duration. Include searching, walking and adjustments.
- 2
Separate internal and external work
Internal work can only happen while the machine is stopped. External work can be done while it runs.
- 3
Convert internal to external
Prepare tools, dies and materials before the stop. Preheat, preassemble and stage everything at the machine.
- 4
Streamline internal work
Use quick-release clamps, standard heights and locating pins. Eliminate adjustments by setting positions in advance.
- 5
Streamline external work
Organize tool storage and carts with 5S. Cut walking and searching during preparation.
- 6
Standardize the new method
Write the changeover as standard work and train every crew. Track changeover time to hold the gain.
Worked example (illustrative)
A molding press stops for 90 minutes per changeover while operators fetch the next mold and its fittings, so it changes product only once a shift. The team stages the mold and a fitting kit on a cart before the stop, preheats the mold and adds locating pins. Changeover falls to about 20 minutes, so the press can change product 4 times a shift and each batch is a quarter of the size, with less stock waiting.
Common mistakes
- Buying faster equipment before separating internal and external work.
- Improving one crew's changeover without writing standard work for all shifts.
- Using the time saved to run even bigger batches instead of smaller, more frequent ones.
When SMED is not the right tool
SMED pays back where changeovers are frequent or long. A machine that rarely changes over gains little.
It also needs a measured baseline. Without timing the current changeover, you cannot see which steps to move or remove.
SMED meaning
SMED stands for single minute exchange of die. The goal is a changeover of under ten minutes, a single-digit number of minutes.
Shigeo Shingo developed the method with Toyota and other Japanese manufacturers. He set it out in his 1985 English book A Revolution in Manufacturing: The SMED System.
SMED steps
Shingo described the method in three main stages. Separate: sort each task into internal, done while the machine is stopped, or external, done while it runs, and move the external ones out of the downtime.
Convert: turn internal tasks into external ones where you can, such as preheating a mold or presetting a fixture. Streamline: shorten what remains with quick clamps, parallel work and fewer adjustments, then write the new method into standard work.
Quick changeover example
Take the illustrative press from the worked example above. The changeover takes 90 minutes, and the video shows about 35 minutes spent fetching the mold, fittings and tools. Staging them on a cart before the stop removes that time from the downtime.
Preheating the mold moves another 15 minutes outside the stop. Locating pins and quick couplings then cut alignment and connection time, leaving about 20 minutes. Each change is small, but together they make smaller batches practical.
Measuring a changeover
Changeover time runs from the last good part of one run to the first good part of the next, at full speed. Adjustment and trial runs count.
Video of a real changeover makes it easy to list each element, time it and sort it into internal and external steps.
Beyond presses
The same idea applies to any switch between jobs, such as line clearance in packaging or turning around an operating room.
Shigeo Shingo set out the method in his book on the SMED system. The time study category covers tools for timing the steps.
Sorting changeover tasks: a first pass
Classify each observed task as a starting hypothesis, then confirm it with the people responsible for the equipment:
| Task | First call | Question to ask |
|---|---|---|
| Collecting the next job's tools | External candidate | Can the full kit be checked before the stop? |
| Removing the installed fixture | Internal | Can removal be simplified without weakening the fixture? |
| Adjusting the new fixture | Conversion candidate | Can an approved preset fixture be prepared offline? |
| Finding the setup sheet | External candidate | Can the current revision be at the machine before the stop? |
| Making and approving the first part | Startup and verification | Which checks need the installed setup and real output? |
Collecting the next job's tools
External candidate
Question to ask: Can the full kit be checked before the stop?
Removing the installed fixture
Internal
Question to ask: Can removal be simplified without weakening the fixture?
Adjusting the new fixture
Conversion candidate
Question to ask: Can an approved preset fixture be prepared offline?
Finding the setup sheet
External candidate
Question to ask: Can the current revision be at the machine before the stop?
Making and approving the first part
Startup and verification
Question to ask: Which checks need the installed setup and real output?
External does not mean outsourced or free; it means the task does not need the machine stopped. A task inside a guarded or hazardous area stays internal even if the motor could technically keep running.
Separating and converting are different moves. Fetching tools before the stop changes when a task happens. Preparing a replacement assembly on a second fixture changes how the work is done, so less is left for the stop.
What to record when observing
Watch a real changeover, not the instruction sheet. For each task, note:
- Task and sequence: A specific description, such as walk to tool store and find wrench, not prepare tools.
- Start and finish time: So durations add up to the total.
- Machine status: Running or stopped.
- Who: The person or role doing it.
- What it needs: Tools, materials, information and approvals.
- Delays and ideas: Why it waited and a first improvement idea.
Video helps where site policy and consent allow. Observe more than one changeover if operators say the first was unusual.
SMED techniques matched to problems
Pick the technique from the problem you observed, not from a checklist:
| Technique | Problem it targets | Control to add |
|---|---|---|
| Changeover cart | Tools and consumables are scattered | A checked inventory and a refill owner |
| Standard fasteners | Operators keep switching tools | Approved fastening specs and good access |
| Fixed stops and locators | Positioning takes repeated trials | Verification of the position and the first-part result |
| Preset tooling | Adjustment uses up stopped time | Tool identification and a reliable presetting method |
| Visual setup references | Settings are hard to find | Product-specific references under revision control |
| Parallel tasks | Independent work done one after another | Work zones, dependencies and a handoff signal |
Changeover cart
Tools and consumables are scattered
Control to add: A checked inventory and a refill owner
Standard fasteners
Operators keep switching tools
Control to add: Approved fastening specs and good access
Fixed stops and locators
Positioning takes repeated trials
Control to add: Verification of the position and the first-part result
Preset tooling
Adjustment uses up stopped time
Control to add: Tool identification and a reliable presetting method
Visual setup references
Settings are hard to find
Control to add: Product-specific references under revision control
Parallel tasks
Independent work done one after another
Control to add: Work zones, dependencies and a handoff signal
Do not remove an adjustment without knowing why it exists. Some variation should be fixed upstream; some needs a legitimate setup adjustment. Ask engineering to tell the two apart.
A worked changeover example
An illustrative changeover takes 40 minutes: 10 minutes collecting tools and information, 18 minutes removing and installing, and 12 minutes adjusting and verifying.
Baseline
Collect tools and information before the stop
Approved presetting removes 6 minutes of adjustment
reduction = (baseline − improved) ÷ baseline × 100
(40 − 24) ÷ 40 × 100 = 40%. That is not single-digit yet, but it points to the next study: the 18 minutes of removal and installation. Note that the first change moved work, while the second removed it. Measure the preparation labor too.
SMED in different processes
Illustrative starting questions by process:
Machining
Can tools, fixtures, programs and setup data be checked before the current job ends? Keep tool verification and first-part checks explicit.
Packaging
Can change parts, labels, materials and instructions for the next format be kitted? Keep the staging area clear so kits are not mixed up.
Molding
How are the next tool and its connections prepared? Treat lifting, handling and temperature changes as engineering questions.
Printing
When do the next job's artwork, plates, materials and settings become available? Separate missing-information delays from real adjustments.
Assembly
Can preconfigured fixture modules and labeled kits replace in-place setup? Verify each module gives the required result once fitted.
Measuring SMED fairly
Answer three questions: did downtime fall, does good production resume reliably, and did the change create work elsewhere?
| Measure | Definition | Watch for |
|---|---|---|
| Changeover duration | Last good part to first good part | The agreed start and end events for each changeover |
| Preparation labor | Time spent before the stop | Work moved, not removed |
| Internal labor | People and time during the stop | Two people for 10 minutes is 20 person-minutes |
| Startup quality | Rejects, repeat checks and corrections | A short stop followed by a long ramp-up |
| Time to stable running | Reaching the agreed operating condition | First good part that is not normal production |
| Readiness failures | Missing tools, materials, instructions or approvals | Kits that are complete only on day shift |
Changeover duration
Last good part to first good part
Watch for: The agreed start and end events for each changeover
Preparation labor
Time spent before the stop
Watch for: Work moved, not removed
Internal labor
People and time during the stop
Watch for: Two people for 10 minutes is 20 person-minutes
Startup quality
Rejects, repeat checks and corrections
Watch for: A short stop followed by a long ramp-up
Time to stable running
Reaching the agreed operating condition
Watch for: First good part that is not normal production
Readiness failures
Missing tools, materials, instructions or approvals
Watch for: Kits that are complete only on day shift
Show the number of changeovers observed and their range, not only the average. Compare like-for-like product pairs, and note any extra technician or unusually good conditions in a trial.
Turning saved minutes into results
Link the expected benefit to a decision the operation will actually make:
- Recovered time: Which demand, backlog or requirement will use it?
- Smaller batches: Which products can be made more often without overloading shared resources? Heijunka uses this.
- Steadier schedules: Is the spread of changeover times narrowing, not only the average?
- Consistent execution: Can every shift follow the same preparation and startup method?
- Less startup waste: Are rejects and repeat adjustments falling?
Recovered minutes are not sales or savings until something changes: if the machine already has spare capacity, the value may be scheduling flexibility. Do not count lower inventory until replenishment rules change and stock is measured.
SMED mistakes and fixes
- 1
A target without a method. Finish faster is not a plan. Go back to the task list and name what will change.
- 2
Preparation without an owner. Give every external task an owner, a time and a visible ready condition.
- 3
Unverified engineering changes. Confirm new clamps, stops, presets or modules do their job.
- 4
A fast trial with hidden help. Record extra people and special preparation that normal production will not have.
- 5
One shift trained. Train and coach every crew that changes over.
- 6
No follow-through. Give open actions an owner and a review date, and check that the promised benefit appears.
Guides on this topic
- One-piece flow: how single-unit flow cuts waiting and WIPOne-piece flow explained: a worked comparison with batches, takt and cycle time, the benefits to check, production and transfer batches, flow and pull, where it fits, the conditions it needs, a pilot plan and how to measure it.Read the guide
- How to measure OEE: start tracking without a big IT projectHow to measure OEE with a free paper tracking sheet before buying software: what to record, how to calculate it and when to automate OEE tracking.Read the guide
- Kaizen event: meaning, steps and a 3-5 day agendaA kaizen event is a focused workshop that improves one process in 3-5 days. Learn the roles, agenda, steps and measures, with a free kaizen charter template.Read the guide
- Lean vs Six Sigma: differences, history and when to use eachLean vs Six Sigma compared: what each method targets, where they came from, how they differ in manufacturing and how Lean Six Sigma combines them.Read the guide
Sources
Books and papers show their publication year. Web pages show the month we last checked them.
- 1.Shigeo Shingo, A Revolution in Manufacturing: The SMED System (Productivity Press, 1985)