The
Eight-Minute Dream That Became an Eight-Hour Reality
Shigeo Shingo walked into a Toyota body-stamping plant in 1969 and
asked a simple question: why does it take four hours to change the dies
on a 1,000-ton press? The engineers told him it was impossible to do it
faster. The operators told him it had always been done this way. The
managers told him the current changeover time was “acceptable.”
By 1970, Shingo and the Toyota team had reduced that changeover from
four hours to three minutes.
This wasn’t magic. It wasn’t new technology. It wasn’t a bigger
budget or more people. It was a systematic rethinking of what actually
happens during a changeover — and more importantly, what
doesn’t need to happen. Shingo called it Single-Minute Exchange
of Die, or SMED, and it became one of the most powerful tools in the
lean manufacturing arsenal.
Fifty-six years later, SMED is taught in every lean course,
referenced in every operations textbook, and featured in every
continuous improvement program. And in the vast majority of
manufacturing facilities around the world, changeovers still take hours
when they could take minutes.
Why? Because most organizations don’t actually do SMED. They do
something that looks like SMED. They buy the concept, borrow
the vocabulary, schedule the workshops, and then quietly revert to the
same changeover process they’ve always had — only now they call it
“optimized.”
The result is one of the most common and least discussed failures in
modern manufacturing: the SMED program that produces no actual
changeover improvement.
What
SMED Actually Is (Before We Talk About How It Fails)
Before examining how organizations sabotage their changeover
reduction efforts, let’s be precise about what SMED means.
SMED is a methodology for dramatically reducing the time required to
switch a manufacturing process from producing one product to producing
another. The “single-minute” in the name refers to the target:
changeovers completed in single-digit minutes (under ten). The core
insight is deceptively simple:
There are two types of changeover activities, and most
organizations never separate them.
Internal activities are tasks that can only be
performed while the machine is stopped. Removing the old die, installing
the new die, aligning it, testing the first part — these require the
equipment to be idle.
External activities are tasks that can be performed
while the machine is still running the current product. Locating the new
die, transporting it to the machine, pre-heating it, gathering tools,
preparing bolts and clamps, pre-setting parameters — all of these can
happen before the machine stops.
The fundamental principle of SMED is this: move as many
internal activities as possible to external time. When the
machine stops, the only things that should happen are the physical
activities that absolutely require the machine to be stopped. Everything
else should already be done.
Then, in a second phase, you systematically eliminate, simplify, or
automate the remaining internal activities. Standardize the bolts so you
need one wrench instead of seven. Use quick-clamp mechanisms instead of
threaded fasteners. Eliminate alignment adjustments by using locating
pins. Pre-set the die height so no shimming is needed.
This sounds straightforward. It is straightforward. And almost nobody
does it correctly.
Failure Mode
#1: The Workshop That Produced a Poster
The most common pattern in failed SMED implementation is the
kaizen event model: a three-day workshop where a
cross-functional team records the current changeover on video,
identifies internal and external activities, brainstorms improvements,
and implements a few quick fixes.
On paper, this is excellent. The team typically reduces changeover
time by 30-50% during the workshop. Everyone celebrates. Photos are
taken. A report is written. The results are presented to leadership.
Then everyone goes back to their normal jobs, and within six weeks,
the changeover time creeps back up to where it was before. Within three
months, it exceeds the original time, because the quick fixes were never
standardized, the pre-changeover preparation tasks were never integrated
into the daily schedule, and the operators who participated in the
workshop have been reassigned.
Here’s what actually happened: the workshop identified external
activities that should be done before the changeover — but
nobody changed the production schedule to create time for those
activities. So the operators continued doing them during the changeover,
because the schedule didn’t give them time to do it any other way.
The poster from the workshop hangs on the wall. The changeover time
is the same as it ever was.
The lesson: SMED is not a workshop. It’s a system. A
three-day event can identify the improvements, but sustaining them
requires changes to scheduling, tool management, die storage, operator
training, and performance measurement. If you’re not willing to change
the system, the workshop is theater.
Failure
Mode #2: Internal Activities Disguised as External
The second most common failure is more subtle, and it traps even
experienced lean practitioners.
During the SMED analysis, the team categorizes each changeover task
as internal or external. They do this by observing the current
changeover and asking: “Does this require the machine to be
stopped?”
The problem is that the current changeover process contains
decades of accumulated bad habits. Tasks that are currently performed
while the machine is stopped may not actually require the
machine to be stopped — they’re just done that way because nobody ever
questioned it.
Example: an operator spends twelve minutes during changeover walking
to the tool crib, finding the right wrench, walking to the supply area,
finding the correct bolts, and returning to the machine. The SMED team
categorizes this as “internal” because the operator does it while the
machine is stopped. But there’s nothing about fetching tools that
requires the machine to be stopped. It was internal only because nobody
prepared the tools in advance.
This sounds obvious, but it’s the single most common SMED error.
Teams take the existing changeover sequence, label each step, and then
try to optimize within those labels. But if the labels are wrong — if
“internal” tasks are actually external tasks that were never pre-staged
— then the entire analysis is built on a flawed foundation.
The fix: Before categorizing tasks, ask a different
question: “Could this task be completed before the machine stops, if
we reorganized the preparation?” If the answer is yes, it’s an
external task that’s currently being performed as internal. The goal
isn’t to categorize what is — it’s to categorize what could
be.
Failure Mode #3: The
Unstandardized Die
SMED’s most famous technical improvement is the elimination of
adjustment. Shingo’s principle was clear: eliminate the need for
adjustment, don’t just speed it up.
In most changeovers, a significant portion of time is spent on
adjustment: aligning the die, setting the height, testing, re-aligning,
testing again, shimming, re-testing. Shingo called these iterative
adjustments the biggest waste in changeover, and his solution was
mechanical: use locating pins, standardized die heights, and quick-clamp
systems so that the die drops into the correct position with zero
adjustment needed.
But standardization is boring. It doesn’t look innovative. It
requires precision machining of die components, consistent maintenance
of locating surfaces, and disciplined storage of dies to prevent damage
to reference points. So organizations skip it.
Instead, they buy quick-change tooling systems that reduce the
fastening time (the time to bolt and unbolt the die) but don’t
address the adjustment time (the time to align, shim, and
test). The result: the bolts come off faster, but the operator still
spends twenty minutes tapping the die into position with a dead-blow
hammer.
The lesson: The adjustment time is the changeover
time. Fastening improvements are meaningful but secondary. If your
changeover still involves trial-and-error alignment, you haven’t done
SMED — you’ve bought fancy clamps.
Failure
Mode #4: The First-Piece Inspection Bottleneck
A changeover isn’t complete until the first good part is produced.
Many organizations reduce the mechanical changeover time — die removal,
installation, alignment — but ignore the qualification time: the period
between “machine is set up” and “first conforming part is approved.”
This qualification phase can take longer than the mechanical
changeover itself. The operator runs a test part. The quality technician
measures it. It’s out of tolerance. The operator adjusts. Another test
part. The technician measures again. Still out. Another adjustment. A
third test part. Finally, it passes.
Each iteration takes fifteen to thirty minutes, depending on cycle
time, measurement complexity, and how long it takes the quality
technician to arrive at the machine (they’re usually covering multiple
lines).
True SMED addresses this by:
- Pre-setting process parameters so the first part is
close to nominal, not just “close enough to test” - Bringing the measurement to the machine — gauges,
fixtures, and instruments at the point of use - Training operators to perform first-piece
inspection instead of waiting for a quality technician - Reducing the number of adjustments through the
mechanical standardization described above
If your mechanical changeover is ten minutes but your first-piece
approval takes forty-five, your SMED implementation has moved the
bottleneck, not eliminated it.
Failure Mode #5: No
Changeover Scheduling
Here’s a question that separates real SMED implementations from
theater: Does your production schedule include time for external
changeover preparation?
In most facilities, the schedule says: “Run Product A until 2:00 PM.
Start changeover at 2:00 PM. Run Product B at 2:30 PM.” The changeover
has thirty minutes.
But the external preparation — finding the die, pre-heating it,
gathering tools, pre-setting parameters — isn’t scheduled. It’s expected
to happen “when someone gets a chance.” Which means it doesn’t happen
until 2:00 PM, when the machine stops and the operator starts doing all
the preparation tasks during the changeover window.
A proper SMED schedule looks different. It includes preparation tasks
as scheduled activities before the changeover: “Operator begins
pre-changeover preparation at 1:30 PM (while Product A is still
running). Machine stops at 2:00 PM. Changeover completed by 2:10 PM.
Product B running by 2:10 PM.”
This requires:
- The scheduler to know what preparation tasks exist and how long they
take - The operator to have time allocated for preparation (which means
someone is watching the running machine while the operator
prepares) - The materials, tools, and dies to be staged at the machine
before the scheduled preparation time
Without this scheduling integration, SMED is a technique that works
in the workshop and fails on the shop floor. The mechanical improvements
are real, but the organizational system doesn’t allow them to
function.
This is why most SMED implementations fail. Not
because the technical improvements are wrong, but because the scheduling
system, the organizational roles, and the daily routines were never
updated to support the new way of changing over.
Failure Mode #6:
Changeovers Are Not Measured
If you ask a plant manager, “What’s your average changeover time for
Press #3?” and the answer is “about thirty minutes, I think” — SMED will
fail in that plant. Guaranteed.
What gets measured gets managed. What gets estimated gets
ignored.
Effective SMED requires:
- Every changeover is timed — start to finish, with
sub-steps recorded - The data is visible — on a board at the machine,
not in a spreadsheet on someone’s laptop - Trends are tracked — is changeover time improving,
stable, or degrading? - Variation is analyzed — why was Tuesday’s
changeover fifteen minutes and Thursday’s thirty-five?
Without this data, there’s no baseline, no way to measure
improvement, and no way to detect regression. The SMED workshop produces
a temporary improvement, and because nobody measures it, nobody notices
when it disappears.
The measurement doesn’t need to be sophisticated. A clipboard at the
machine where operators log start time, die-out time, die-in time,
first-good-part time — that’s enough. The discipline of measurement
drives the discipline of improvement.
What Real SMED Looks Like
When SMED is done correctly — truly, systemically, sustainably — the
results are transformative:
- Changeover times drop by 70-90%, not in a workshop
but permanently - Production flexibility increases because smaller
batch sizes become economically viable (the changeover cost is no longer
the constraint) - Inventory decreases because you can run smaller
batches more frequently without paying the changeover penalty - Operator morale improves because changeovers become
structured, predictable events instead of chaotic scrambles - Quality improves because the first part is closer
to nominal, reducing the scrap and rework from the qualification
phase
But achieving these results requires more than a workshop and a
poster. It requires:
- Sustained leadership commitment — not just
“support” but active participation in removing the organizational
barriers (scheduling, roles, responsibilities) - Mechanical standardization — dies, tools, and
clamping systems engineered for zero-adjustment changeover - Scheduling integration — preparation tasks built
into the production schedule, not hoped for - Daily measurement — changeover time tracked,
visible, and discussed - Operator ownership — the people who do the
changeover are the people who improve the changeover - Continuous refinement — SMED isn’t a project; it’s
a practice. Every changeover is an opportunity to find another second to
eliminate.
The Uncomfortable Question
If your organization has “implemented SMED,” ask yourself
honestly:
- What was your changeover time before SMED? What is it now?
- If you stopped all SMED-related activities tomorrow, would the
changeover time change? - Are external preparation tasks actually performed while the machine
is running — or is that what the procedure says and nobody follows? - When was the last time someone questioned whether an “internal” task
could be moved to external? - Does your production scheduler know what external preparation tasks
exist?
If the answers make you uncomfortable, you’re in good company. Most
organizations are in the same position. The gap between knowing SMED and
doing SMED is enormous — and it’s where nearly every manufacturer
fails.
Shingo didn’t give us a complicated tool. He gave us a simple
principle: separate what must stop the machine from what doesn’t, then
eliminate what does. The principle hasn’t changed in fifty-six years.
The execution is still the hard part.
Stop celebrating the workshop. Start measuring the changeover. That’s
where SMED begins.
About the Author: Peter Stasko is a Quality
Architect with over 25 years of experience in manufacturing quality
management, lean implementation, and continuous improvement. He has led
quality transformations across automotive, electronics, and industrial
manufacturing sectors, specializing in bridging the gap between
theoretical quality frameworks and practical shop-floor execution.
