The Promise That Sold Itself
Shigeo Shingo walked into a manufacturing plant in 1969 and did
something that sounded impossible. He reduced a stamping press setup
from four hours to three minutes. Not by buying new equipment. Not by
hiring faster operators. By thinking differently about what “setup”
actually meant.
The insight was elegant: separate internal setup (activities that
require the machine to be stopped) from external setup (activities that
can be done while the machine is still running), then convert as much
internal setup to external as possible. What remained as internal setup
could be streamlined with parallel operations, standardized clamping,
and elimination of adjustments.
The manufacturing world adopted SMED with the enthusiasm of a
converts’ movement. Consultants built careers on it. Books were written.
Certification programs sprouted. And somewhere along the way, the
revolutionary simplicity of Shingo’s insight got buried under thirty
years of implementation debris.
Today, in factories across the world, SMED is spoken about in hushed,
reverential tones during audits and management reviews. Setup reduction
charts hang on walls. Kaizen events are dedicated to shaving seconds off
changeovers. And the actual changeover times — the ones that matter, the
ones that determine whether your factory is flexible or rigid, whether
you can run small batches economically or whether you’re trapped in long
production runs building inventory nobody ordered — those changeover
times haven’t meaningfully improved in years.
This is the story of how a methodology designed to eliminate waste in
setup became a methodology that generates waste of its own.
What
SMED Actually Is (Before We Talk About What It Becomes)
Let’s give SMED its due. The core methodology is sound, and when
applied with genuine understanding, it works.
The framework operates in four stages:
Stage 1: Separate internal and external setup. This
alone is profound. Most factories conflate the two. Operators stop the
machine, then go find the dies, then go find the tools, then go find the
bolts, then look for the crane operator. Each of these steps could have
been done while the machine was still running the previous job. Shingo’s
first insight is simply this: if you can do it while the machine runs,
do it while the machine runs.
Stage 2: Convert internal setup to external setup.
Once you can see which steps require the machine stopped, you attack
them. Can pre-assembly of dies happen offline? Can tools be staged on a
cart before the changeover begins? Can temperature conditioning of molds
happen in a preheat station while the current job finishes?
Stage 3: Streamline remaining internal setup.
Whatever internal setup remains, make it faster. Parallel operations
(two operators working simultaneously instead of one waiting). Quick
clamps instead of bolts. Elimination of adjustments through standardized
positioning. This is where the engineering gets creative.
Stage 4: Eliminate adjustments entirely. The
ultimate goal. Every adjustment during setup is an admission that
something isn’t positioned correctly the first time. If you can design
the setup so that components click into a definitive position with zero
adjustment needed, you’ve achieved what SMED promises.
This is good methodology. It works. So what goes wrong?
The First
Failure Mode: SMED as an Event, Not a System
The most common failure pattern in SMED implementation is the
kaizen-event approach. Management decides setup times are too long. A
cross-functional team is assembled. A SMED workshop is scheduled for
three days. Consultants are brought in. Video cameras are set up. The
current changeover is filmed, analyzed, dissected. Improvement ideas
flow. Quick clamps are ordered. Carts are built. Standard work is
documented.
At the end of the three days, the team celebrates. Setup time has
been cut from 90 minutes to 35 minutes. Charts are updated.
Presentations are made to leadership. Everyone feels good.
Six months later, the setup time is back to 75 minutes.
What happened? The SMED event solved the technical problem but not
the systemic one. The quick clamps worked, but when one broke, nobody
ordered a replacement, and the operator went back to using bolts. The
tool cart was beautifully organized during the event, but over time it
became a dumping ground for everything. The standard work document was
laminated and posted at the workstation — and then a new product was
introduced that required a different setup sequence, and nobody updated
it.
The event created a temporary improvement. But improvement is not a
destination; it’s a discipline. And SMED, as practiced in most
organizations, is an event, not a discipline. The methodology becomes a
thing you did, not a thing you do. The improvement
decays because nothing in the system sustains it.
This is the first irony of SMED: a methodology whose entire premise
is converting temporary setup states into permanent efficiencies
becomes, itself, a temporary efficiency that decays back into permanent
inefficiency.
The
Second Failure Mode: Optimizing the Wrong Changeover
Here’s a scenario I’ve seen more times than I can count.
A factory runs a SMED kaizen event on Press Line A. Press Line A does
a changeover twice per week. The team successfully reduces the
changeover from 120 minutes to 60 minutes. That’s 120 minutes saved per
week — two hours reclaimed for production.
Meanwhile, Press Line B, in the same factory, does fourteen
changeovers per week at 90 minutes each. That’s 1,260 minutes of
changeover per week. Nobody touches Press Line B because it wasn’t the
one selected for the kaizen event.
The factory reports a “50% reduction in setup time” in its monthly
metrics. Technically true for Press Line A. Practically meaningless for
the overall plant capacity.
This is the second failure mode: SMED optimization applied to the
wrong constraint. Organizations pick the changeover that’s most visible,
most annoying, or most politically convenient — not the one that matters
most for throughput. The Theory of Constraints tells us that improvement
anywhere except the bottleneck is an illusion. But SMED events are
rarely selected based on constraint analysis. They’re selected based on
who volunteered, which line has the most vocal supervisor, or which area
looks worst on the gemba walk.
The result: you get really fast at changing over a machine that
doesn’t determine your factory’s output, while the machine that actually
limits your throughput continues to bleed hours every week in
inefficient setups that nobody studies.
The
Third Failure Mode: The External Setup That Isn’t External
Stage 1 of SMED says: separate internal and external setup. Do the
external work while the machine is still running.
This sounds straightforward. But in practice, it requires a level of
preparation discipline that most factories simply don’t have. Let me
describe what actually happens.
The operator knows the next job requires Die Set 47. The SMED
standard work says: stage Die Set 47 on the preparation cart, verify all
components are present, preheat the die to operating temperature, and
position the cart at the workstation — all before the current job
ends.
In reality: the operator finishes the current job, stops the machine,
and then goes to find Die Set 47. Die Set 47 is in the tool
crib, but the crib attendant is on break. When the attendant returns,
Die Set 47 is missing a locating pin. The operator hunts down a
replacement pin. The die needs preheating, but the preheat station is
being used for a different job. Twenty minutes have passed before the
operator even brings the die to the machine.
Every minute of this is supposed to be external setup — done while
the machine was still running the previous job. But the standard work
for external setup exists only on paper. In practice, nobody does
pre-changeover preparation because nobody assigned responsibility for
it, nobody provided the time for it, and nobody audits whether it
actually happens.
The SMED documentation shows a 25-minute internal setup time. The
actual changeover takes 55 minutes because 30 minutes of “external” work
happens after the machine stops. But the standard work says external, so
the metrics report 25 minutes. The gap between documented changeover
time and actual changeover time is where factories hide their
inefficiency, and SMED provides the perfect cover.
The
Fourth Failure Mode: Quick Clamps That Aren’t Quick
The hardware side of SMED is seductive. Quick-release clamps,
hydraulic locking systems, pre-positioned locating pins, color-coded
tooling. When you visit a factory that’s implemented SMED, the hardware
looks impressive. Everything has a lever or a cam or a pneumatic
actuator.
But here’s what the brochures don’t tell you: quick clamps are quick
only when everything is perfectly aligned. The moment there’s a
dimensional variation in the die, a worn locating surface, or thermal
expansion from a die that wasn’t properly preheated, the quick clamp
doesn’t engage. The operator grabs a dead-blow hammer. Then a bigger
hammer. Then a pry bar. The “quick” changeover becomes a wrestling match
between one human and several hundred kilograms of tooling.
And once the clamp is forced into position with a hammer, the
alignment is compromised. The first part produced after the changeover
is out of tolerance. The operator makes adjustments — exactly what Stage
4 of SMED says to eliminate. But the adjustments are necessary because
the setup conditions were forced, not engineered.
This creates a vicious cycle: the quick clamp system, which was
supposed to eliminate adjustment, becomes the cause of
adjustment. And because the clamps are labeled “quick” and documented in
the SMED standard work as requiring zero adjustment, the time spent
adjusting is reported as something else. Setup time looks fine on paper.
First-pass yield tells a different story.
The
Fifth Failure Mode: SMED for Products That Don’t Need Flexibility
SMED’s value proposition is flexibility. Shorter changeovers mean you
can run smaller batches economically, which means you can be more
responsive to customer demand, carry less finished-goods inventory, and
reduce obsolescence risk.
This makes sense for high-mix, make-to-order environments. It makes
less sense for high-volume, dedicated production lines that run the same
product for months at a time.
Yet I’ve seen factories invest significant resources in SMED for
lines that do two changeovers per year. The setup takes eight hours. A
SMED event reduces it to four hours. The factory saves eight hours per
year — one production shift — on a line that runs 5,000 shifts per year.
The ROI of the SMED investment, if anyone bothered to calculate it, is
negative.
But SMED has become a quality-program checkbox. “Have you implemented
setup reduction?” is on every operational-excellence audit. The answer
needs to be yes. Whether the improvement matters doesn’t enter the
conversation. The methodology has become an end in itself — something
you implement because the audit asks for it, not because the business
benefits from it.
This is perhaps the most insidious failure mode: SMED as compliance
ritual. You do it because someone checks whether you did it. The actual
improvement is irrelevant. The documentation is the deliverable.
The Sixth Failure
Mode: The Setup Time Paradox
Here’s a pattern that’s counterintuitive but remarkably common.
A factory implements SMED and reduces changeover time from 120
minutes to 30 minutes. This is a genuine, sustained improvement.
External setup is done properly. Quick clamps work. Standard work is
followed.
In response, production planning increases the number of changeovers.
If changeovers are cheap, run smaller batches. Smaller batches mean less
inventory. Less inventory means lower carrying costs. The math
works.
But now the factory is doing four times as many changeovers. Each one
is 30 minutes instead of 120 minutes. Total changeover time per week:
identical to what it was before SMED. The machine availability hasn’t
improved. The factory’s output capacity hasn’t increased. What changed
is the batch size — and with it, the administrative complexity of
scheduling, material handling, and quality verification for each new
batch.
The factory traded one form of waste (excess inventory) for another
form of waste (excess changeovers and the associated complexity).
Whether this trade is beneficial depends on factors that nobody
analyzed: the cost of carrying inventory versus the cost of additional
scheduling complexity, the quality risk of more frequent changeovers
(each one is an opportunity for a setup error), and the stress placed on
operators who now do four changeovers per shift instead of one.
SMED made changeovers cheap, so the organization optimized for more
of them. But cheap isn’t free. The setup time paradox is that reducing
the cost of changeovers can increase the total cost of changeovers if
the organization simply increases the quantity to compensate. The
methodology solved the unit-cost problem and created the total-cost
problem.
The Seventh
Failure Mode: Digital SMED Theater
We now live in an era of digital manufacturing. Industry 4.0. Smart
factories. IoT-enabled everything. And SMED has not escaped
digitization.
Factories now install cameras above their machines to record
changeovers. AI-powered video analytics identify wasted motion.
Augmented reality headsets guide operators through setup sequences.
Digital dashboards display real-time changeover timers. Automated alerts
fire when a changeover exceeds the target time.
All of this technology generates data. The data generates reports.
The reports generate meetings. The meetings generate action items. The
action items generate… more data in the next cycle.
What the technology doesn’t generate is actual changeover
improvement. Because the bottleneck was never information. The operators
always knew where the wasted time was. They knew the locating pin was
worn. They knew the preheat station was undersized. They knew the tool
cart was disorganized. They told their supervisors six months ago. The
supervisors told their managers. The managers asked for data. Now they
have data. The pin is still worn. The preheat station is still
undersized. The cart is still disorganized.
Digital SMED theater is the most expensive form of the problem. It
costs hundreds of thousands of dollars in sensors, software, and
dashboards to confirm what the operators already knew — and then it
changes nothing, because the underlying issues require investment in
hardware, training, and organizational discipline, not investment in
monitoring systems.
The factory can now watch its slow changeovers in high-definition
video, with AI-generated annotations explaining exactly why they’re
slow, in real time, on a dashboard that emails a PDF summary to
management every Monday morning.
Shingo would not be impressed.
How to Actually Make SMED
Work
Despite all of this, SMED is not a failed methodology. It’s a
methodology that fails when it’s applied as a program rather than as a
principle. Here’s what genuine SMED practice looks like:
Focus on the constraint. Before optimizing any
changeover, identify the bottleneck. Use Value Stream Mapping or
throughput analysis to find the machine whose changeover time directly
limits factory output. SMED everywhere except the constraint is
waste.
Sustain, don’t just improve. Every SMED improvement
must have a sustainment mechanism. Monthly audits of external setup
compliance. Visual management of tool carts with shadow boards and
missing-item indicators. Preventive maintenance on quick-clamp hardware.
Standard work that’s updated when products change, not just when
auditors visit.
Measure what matters. Stop measuring setup time in
isolation. Measure total changeover cost (setup time × frequency ×
quality risk × scheduling complexity). Measure first-part yield after
changeover, not just the time to complete the changeover. A 15-minute
changeover that produces 30 minutes of scrap parts is worse than a
45-minute changeover that produces good parts immediately.
Make external setup someone’s job. Not the
operator’s job in addition to running the machine. A dedicated setup
technician or a staging team whose performance is measured on changeover
readiness. If external setup is everyone’s responsibility, it’s no one’s
responsibility.
Invest in the boring stuff. The quick clamps get the
attention, but it’s the preheat stations, the tool-crib organization,
the die-storage system, and the preventive maintenance of locating
surfaces that actually determine changeover performance. These aren’t
exciting kaizen-event material. They’re infrastructure. Treat them as
such.
Respect the operator’s knowledge. The operator who
has done a thousand changeovers knows more about setup efficiency than
any consultant, any video analytics system, and any standard work
document. The most effective SMED improvements I’ve seen came from
operators who were asked, listened to, and given the resources to
implement their own ideas. The least effective came from engineers who
watched a video of the changeover and designed a solution from behind a
desk.
The Uncomfortable Question
If your factory has implemented SMED, I want you to ask yourself one
question: What is your actual changeover time right now — not
the documented standard, not the best-ever achieved during the kaizen
event, but the time from last good part of the previous job to first
good part of the next job, measured by someone standing at the machine
with a stopwatch, today?
If that number is more than 20% longer than your documented setup
time, your SMED implementation is theater. The gap between your
documented performance and your actual performance is the gap between
what your quality system claims and what your factory delivers.
Shingo didn’t create a methodology for generating impressive reports.
He created a methodology for making machines more productive. The
methodology still works. The question is whether you’re working the
methodology — or whether the methodology is working you.
About the Author: Peter Stasko is a Quality
Architect with over 25 years of experience in manufacturing quality
management, process optimization, and operational excellence. He has
implemented quality systems across automotive, electronics, and heavy
industries on three continents, and writes about the gap between what
quality methodologies promise and what they actually deliver on the
factory floor.
