What does SMED stand for?

SMED stands for Single-Minute Exchange of Die. "Single-minute" means a single-digit number of minutes, so the goal is any changeover under 10 minutes, not under 60 seconds. The method was developed by Shigeo Shingo and published in English in 1985 as A Revolution in Manufacturing: The SMED System.

How is changeover time measured?

From the last good piece of the previous product to the first good piece of the next product at normal rate and quality. Starting the clock at machine stop or ending it at restart hides ramp-down, trial parts, and adjustment scrap, which are usually the biggest losses.

How much can SMED reduce changeover time?

Shingo's documented projects showed reductions of 90 percent and more, most famously a 1,000-ton Toyota stamping press that went from four hours to about three minutes in 1969. Your first pass, just separating internal from external work, commonly removes a third to a half of the stop time with little or no capital spending.

What is the connection between SMED and batch size?

Long changeovers force long runs and big batches to amortize the lost time. Short changeovers make small batches economical, which enables leveled scheduling (heijunka), pull systems like kanban, lower inventory, and faster response to demand changes. The schedule flexibility, not the saved minutes, is the real prize.

Does SMED apply outside of stamping presses?

Yes. The internal/external logic applies to any switch between products or jobs: packaging line changeovers, CNC job setups, injection-mold swaps, filler and labeler changeovers, even cleaning and sanitation sequences in food plants. Anywhere a line stops to switch, the same separate-convert-streamline sequence works.

How do you keep changeover times from creeping back up?

Standardize the improved method, train it on every shift, and keep measuring every changeover permanently. Times decay when carts stop being kitted and checklists stop being used. Real-time capture at the station, rather than end-of-shift paper logs, lets supervisors see drift the day it starts.

SMED: Cut Changeover Time Without Cutting Corners

Q: What is the difference between internal and external setup?

Internal setup is work that can only be done while the machine is stopped, like physically swapping the die. External setup is work that can be done while the machine is still producing, like staging the next die, kitting tools, and preheating. Moving external work outside the stop window is the first and cheapest SMED gain.

The single-minute exchange of die, from Shingo's stamping presses to your Monday-morning changeover.

SMED (Single-Minute Exchange of Die) is a lean method for cutting changeover time by separating setup work into external tasks done while the machine runs and internal tasks done while it is stopped, then converting and streamlining both until changeovers take single-digit minutes.

Changeover is the tax you pay every time a line switches from one product to the next, and on many floors it is the single biggest chunk of planned downtime. SMED is the discipline that shrinks that tax without shortcuts that come back as quality escapes. It is a core tool in lean manufacturing because short changeovers make small batches economical, and small batches open the door to everything else: less inventory, faster response, and schedules that can actually follow demand. This post covers where SMED came from, the internal/external distinction that does most of the work, an eight-step framework you can run on any machine, and the corner-cutting traps to avoid.

Where Did SMED Come From?

SMED was developed by Japanese industrial engineer Shigeo Shingo over roughly two decades of shop-floor projects. Three of them mark the method's history:

1950, Toyo Kogyo (Mazda), Hiroshima. Studying press changeovers, Shingo noticed a crew losing an hour hunting for a missing bolt while the press sat idle. That observation produced the core distinction: some setup work (finding bolts, staging dies) can happen while the machine runs; only some truly requires the machine stopped.
1957, Mitsubishi Heavy Industries shipyard. Shingo doubled the output of a diesel-engine bed-planer by moving marking and centering work off the machine onto a spare table, so the machine changed over in a fraction of the time.
1969, Toyota Motor Company. Working on a 1,000-ton stamping press whose changeover took four hours, Shingo and the Toyota team first cut it to 90 minutes, then, pushed by management to go further, reached about three minutes. That result gave the method its name: setups in a single-digit number of minutes.

Shingo published the method in English in 1985 as A Revolution in Manufacturing: The SMED System (Productivity Press). The name is worth reading literally: "single-minute" does not mean 60 seconds. It means under 10 minutes, one digit on the clock (background on the method's history).

What Is the Difference Between Internal and External Setup?

Internal setup is any task that can only be done while the machine is stopped: unbolting the old die, mounting the new one, connecting utilities. External setup is any task that can be done while the machine is still producing: retrieving the next die, staging tools and fixtures, preheating, preparing paperwork, kitting fasteners. The distinction matters because on an unimproved changeover, crews routinely do external work during the stop, which means the machine sits idle while somebody walks to the tool crib. Simply sorting tasks into the two buckets and moving external work outside the stop window commonly removes a large share of downtime before you buy a single quick-release clamp.

Separating internal from external setup shrinks the stop window before any hardware changes. External tasks run in parallel with production, before and after the swap.

The 8-Step SMED Framework

Shingo described SMED as three conceptual stages: separate internal from external setup, convert internal to external, and streamline everything that remains. In practice, running it as eight concrete steps on one machine works better for a first project:

Pick one changeover and film it. Choose a high-frequency changeover on a constraint machine. Record the whole thing, last good part to first good part, and note every task, walk, and wait with timestamps.
Classify every task as internal or external. Be strict: a task is internal only if it is physically impossible with the machine running. Most crews find a third or more of the stop is spent on work that never needed the machine stopped.
Move external tasks outside the stop. Stage the next die, tools, fixtures, materials, and paperwork at the machine before the stop starts. Build a setup cart and a pre-changeover checklist so staging is standard, not heroic.
Convert internal tasks to external. Preheat dies outside the press. Pre-assemble tooling on standardized base plates. Duplicate fixtures so the next job is mounted and dialed in while the current job runs.
Streamline the remaining internal work. Replace bolts with quick-release clamps and cam locks, standardize fastener sizes so one tool does the job, use locating pins and stops so position is fixed by hardware, and run tasks in parallel with two people where it is safe.
Eliminate adjustment. Adjustment and trial runs often eat half of a traditional changeover. Replace trial-and-error dial-in with preset gauges, shims, templates, and reference settings recorded from the last good run.
Streamline external work too. Shadow boards, kitted carts, point-of-use storage, and a standard sequence for the pre-stage routine. External time is not free, it is somebody's labor, and sloppy external prep eventually leaks back into the stop.
Standardize, measure, repeat. Write the new method as standard work train every crew and every shift on it, track every changeover against the target, and run the next cycle on the next biggest time block.

An illustrative waterfall, not data from a specific plant: each SMED stage removes a layer of stopped-machine time, with adjustment elimination doing the final push into single digits.

What Does a Pit Stop Teach About Changeover?

A race-car pit stop is the cleanest picture of SMED thinking anywhere outside a factory. Everything that can possibly happen before the car stops has already happened: tires staged and pre-warmed, guns charged, every crew member in position with one defined job. The stop itself contains only work that physically requires the car present, done in parallel, with zero adjustment because the hardware locates itself. Then the crew resets and rehearses. No step was skipped, and nothing about the car's safety was compromised to get the time down. That is the standard to copy: fast because prepared, not fast because careless.

The pit-stop pattern: preparation while running, parallel no-adjustment work during the stop, then reset and rehearsal so the next stop is faster still.

How Do You Measure Changeover Time?

Changeover time runs from the last good piece of the old product to the first good piece of the new product at normal rate and quality. Both endpoints matter. Starting the clock at "machine stopped" hides slow ramp-down, and stopping it at "machine restarted" hides the trial parts, tweaking, and scrap that follow a bad setup. Measured honestly, changeover shows up as setup and adjustment loss in the six big losses and drags the availability term of OEE, which is why plants that track machine downtime by reason code usually find changeover at or near the top of the Pareto.

The capacity context makes the prize concrete. The Federal Reserve's G.17 industrial production report has shown U.S. manufacturing capacity utilization running in the mid-70s percent range in recent years (Federal Reserve, G.17 Industrial Production and Capacity Utilization), meaning a meaningful slice of installed capacity sits unused. Every changeover hour recovered on a constraint machine is capacity you already paid for. Shingo's own published record, from the 1950 Toyo Kogyo project through the 1969 Toyota press that went from four hours to about three minutes, shows reductions of 90 percent and more were achievable with method changes, mostly before spending on new equipment.

How Do You Cut Changeover Time Without Cutting Corners?

The failure mode hiding in every changeover project is speed bought with risk. Guard against these specifically:

Never externalize the first-piece quality check. The changeover is not done until a good part is verified. Teams chasing a stopwatch number sometimes ship the check downstream; that converts downtime into scrap and customer complaints.
Do not delete torque, lockout, or safety steps. If lockout/tagout makes the changeover slower, the answer is better staging around it, never skipping it.
Do not confuse averages with the standard. One hero crew hitting 12 minutes while other shifts take 40 is not a capability, it is tribal knowledge. Film the best method, write it down, train it everywhere.
Watch for decay. Changeover times drift back up when carts stop being kitted and checklists stop being used. Keep the measurement running permanently, not just during the project.

Why Do Faster Changeovers Change the Whole Schedule?

The point of SMED is not saving minutes, it is what short changeovers let you do with the schedule. When a changeover costs four hours, the economics push you to long runs and big batches, which bloat inventory and make the plant slow to respond when demand shifts. When it costs eight minutes, you can run every product more often in smaller batches, which is exactly what heijunka (production leveling) and pull systems like kanban require. Short changeovers are the enabling condition: you cannot level a schedule you cannot afford to switch. They also protect takt time because a line that changes over quickly loses fewer beats to the switch.

What Does a First SMED Project Look Like?

A first project works best as a focused week, not a standing committee. Scope it to one machine and one high-frequency changeover, ideally on the constraint, because time recovered anywhere else is not capacity. The team is small and specific: the operators who actually run the changeover on each shift, a setup or tool-room person, a maintenance tech, and a supervisor who can approve small purchases on the spot. Day one is filming and task listing; day two is classification and the internal/external sort; day three is building the cart, the kits, and the checklist, and trialing the new sequence on a live changeover; day four is a second trial with times captured and the standard written; day five is training the other shifts and setting the tracking. Budget for modest hardware, clamps, pins, duplicate fittings, a cart, and hold the big capital requests until the method gains are banked. Expect the first pass to land well short of single digits; the point of week one is a repeatable standard and a measurement habit, because the second and third passes are where the single-digit changeovers come from.

How Do You Sustain SMED Gains?

The plants that hold their gains treat changeover as a measured, managed event, not a black hole between runs. That means a reason-coded record of every changeover, the standard visible at the machine, and a review cadence when times drift. This is also where digitization earns its keep: paper changeover logs get filled in at end of shift from memory, while tablet-based capture at the station timestamps each changeover as it happens, so supervisors see the drift the day it starts instead of in next month's report. That pattern, replacing paper logging with real-time capture and automated reporting, is exactly what CLS did across its shops (see the CLS case study), and it is the foundation the rest of a changeover program stands on. No rip-and-replace, just honest numbers arriving fast enough to act on.