SMED (Single-Minute Exchange of Dies): The Complete Guide to Quick Changeover

SMED, short for single-minute exchange of dies, is a method for cutting equipment changeover from hours to single-digit minutes by reorganizing the setup work, not by rushing it. Faster changeovers do more than save time. They let you run smaller batches without losing efficiency, which means less inventory, shorter lead times, and far less production lost to switching between products. This guide covers what SMED is, where it came from, the internal-versus-external idea at its core, the step-by-step method and techniques, how to target and run a SMED event, and what it does for OEE.

What SMED is

SMED is a systematic way to reduce the time it takes to change a machine over from making one product to making the next. It is also called quick changeover or quick die change, and the principle applies far beyond the stamping dies it was named for, to any setup or product switch on a line.

The "single-minute" in the name refers to single-digit minutes, a changeover under ten minutes, not a literal sixty seconds. What makes SMED rigorous is how it defines the changeover in the first place. The clock runs from the last good part of the old product to the first good part of the new product at full speed and full quality. Everything in between counts: the walk to the tool crib, the hunt for the right wrench, the first-off inspection, and any adjustments after the restart. That strict definition is what separates real SMED from informal "we tightened up the setup a bit" exercises, because it captures all the hidden time that usually goes unmeasured.

A simple image shows the scale of what is possible. Changing one tire in a driveway might take fifteen minutes, while a racing pit crew changes four in well under fifteen seconds. They manage it by preparing everything before the car stops, working in parallel, and drilling a standardized routine, which are exactly the three moves SMED applies to a production changeover.

A short history of SMED

SMED came out of the Toyota Production System. Taiichi Ohno was trying to build a just-in-time, small-batch system, but changeovers on the big stamping presses could take up to four hours, which made small batches uneconomical and blocked the whole approach. Ohno pushed his team to slash setup times, and the industrial engineer Shigeo Shingo developed the method that did it.

Shingo refined SMED across roughly two decades and three settings, beginning at Toyo Kogyo, now Mazda, in 1950, where he first drew the distinction between internal and external setup, then at Mitsubishi Heavy Industries in 1957, and finally at Toyota in 1969. His most cited result took the changeover on a 1,000-ton press from four hours to three minutes. He published the method for a global audience in his 1985 book, and his documented results across a wide range of companies averaged changeover reductions of around 94 percent.

Why changeover time matters

Changeover time is pure availability loss. While a machine is being set up it is not making anything, which is why setup and adjustments is one of the Six Big Losses that drag down overall equipment effectiveness. Every minute cut from a changeover is a minute of capacity handed back to production.

The deeper effect is on batch size. When a changeover is long, the only way to keep its cost per part low is to run a huge batch and spread the setup across thousands of units. That forces big production runs, which pile up inventory, lengthen lead times, and make the plant slow to respond when demand shifts. Cut the changeover and the math flips: small batches become economical, inventory falls, lead times shrink, and the line can switch to whatever the customer actually needs next. This is why SMED is the technical backbone of just-in-time and every pull system. The flexibility that lean depends on is impossible until changeovers are fast.

Internal vs external setup: the core idea

The single most important move in SMED is sorting setup work into two kinds, usually called elements.

• Internal elements can only be done while the machine is stopped, such as removing the old die and bolting in the new one.
• External elements can be done while the machine is still running, such as gathering and staging tools, preheating a die, kitting the next batch's materials, and verifying settings in advance.

Most plants do far more work during downtime than they need to, simply out of habit. In a typical unimproved changeover, a large share of what operators do while the machine is stopped is actually external work that was just postponed, and that share is recoverable downtime. Shingo found that separating the two and moving external work off the clock often cuts changeover time by 30 to 50 percent before any clever engineering is involved.

Here is the distinction in concrete terms.

Internal (machine stopped)	External (machine running)
Removing and installing the die or tooling	Retrieving the next die and staging it at the press
Cleaning and aligning the fixture	Preheating the die to operating temperature
Connecting and disconnecting services	Kitting the next run's raw material and consumables
Confirming first-off dimensions	Laying out tools on a staged cart in sequence

The SMED method, step by step

Shingo's method, with the modern habit of measuring first, comes down to four steps. They look sequential but in practice you cycle through them.

1. Measure the current changeover. Record an actual changeover, ideally on video, and break it into individual elements with a time for each. You cannot improve what you have not seen in detail, and operators are almost always surprised by where the time really goes.
2. Separate internal and external setup. Go through every element and classify it. Then make sure all the external work is actually done before the line stops or after it restarts, not during the downtime. This step alone delivers the first large gain.
3. Convert internal to external. Find ways to move internal work outside the stop: preheat dies so they do not need warming in place, pre-stage and pre-assemble tooling, use duplicate or modular fixtures so the next setup is ready to drop in, and prepare and verify everything in advance.
4. Streamline every element. Attack the time of the remaining internal and external work with the techniques below, so the stopped time keeps shrinking and the prep gets leaner too.

SMED techniques

Most of the streamlining gains come from a familiar toolbox.

• Quick-release clamps and functional fasteners. Bolts are slow, and a bolt is only doing its job on the last turn. Replace them with quick-action clamps, cam locks, or one-turn methods so a fastener engages in a single motion.
• Eliminate adjustments. Adjustment and trial-and-error after the restart is often the single biggest hidden cost in a changeover. Replace it with fixed stops, gauges, shims, visible centerlines, and standardized numerical settings so the setup lands in the right position by design. The goal is to set, not to adjust.
• Parallel operations. Two people working different sides of a machine at once can roughly halve the stopped time, as long as the steps are choreographed and safe.
• Pre-staging and pre-heating. Stage the next die, tools, and materials on a cart in the order they are used, with a shadow board so nothing is hunted for. Bring anything that needs to be hot up to temperature while the line still runs.
• Standardize the changeover. Document the changeover as standard work with a clear sequence and checklist, so it is done the same fast way on every shift rather than reinvented each time.
• Mechanize last. Hydraulic clamping and automated die changers have their place, but only after the cheap organizational wins are captured. Mechanizing a wasteful changeover just automates the waste.

Start with the human elements

It helps to split changeover improvements into two kinds. Human improvements come from preparation, organization, and sequencing: staging tools, separating internal from external work, writing a standard, choreographing who does what. Technical improvements come from engineering: new clamps, modified tooling, mechanization. The human improvements are almost always faster and cheaper, and they usually deliver the first large drop in changeover time. The common mistake, especially on engineering-minded teams, is to jump straight to technical fixes and spend money before the free organizational wins have been taken. Capture the human gains first, then engineer what is left.

OTED and the limits of single-minute

Once a team reaches single-digit minutes, the next target is one-touch exchange of die, or OTED, a changeover in under 100 seconds. Beyond that is no-touch exchange, where the changeover is fully automated and needs no operator intervention at all. Few changeovers need to go that far, but the progression is a useful reminder that the single-minute goal is a milestone, not a ceiling.

Before you start: is SMED your priority?

SMED is worth doing only where changeover is actually a major source of lost time, and the way to know is to measure rather than assume. Before launching a program, put a system in place to see where productive time goes, using OEE broken down into the Six Big Losses, with availability loss split into downtime reasons that include a code for changeover. Collect a couple of weeks of real data.

If changeover turns out to be a significant share of lost time, a common rule of thumb being around 20 percent or more, SMED is a strong bet. If most of your losses come from unplanned breakdowns or slow cycles instead, the bigger prize is elsewhere and SMED can wait. Targeting the right loss first is the difference between an improvement project that moves the number and one that polishes a corner nobody was waiting on.

How to run a SMED event

In practice, improving a changeover is usually run as a focused event on one machine.

1. Pick a pilot. Choose a changeover that hurts: a frequent one, a long one, or one on a bottleneck where lost time is most expensive. A changeover of around an hour with high variation in its times is an ideal candidate, with enough room to improve but not so much that the scope overwhelms the team.
2. Record and break it down. Film a real changeover and list every element with its duration, working with the operators who run it rather than around them. A typical changeover breaks into 30 to 50 elements. Capture both the human elements, what the operator does, and the machine elements, what the equipment does, since the human ones are usually the easiest to cut. Sticky notes laid out on a wall in sequence are a fast way to map them, and a second observer will catch what the camera misses. Watch for the Hawthorne effect, where a changeover speeds up simply because it is being watched, so baseline from historical data where you can.
3. Classify and challenge. Mark each element internal or external, then work through the room asking what can move external, what can be eliminated, and what can be streamlined, using a quick cost-versus-benefit judgment to decide what to tackle first.
4. Implement and standardize. Make the changes, build the staging carts and quick clamps, write the new standard, and train the team to it.
5. Measure and repeat. Time the improved changeover, confirm the gain, then move to the next machine with the same playbook.

The single biggest predictor of whether SMED sticks is whether the frontline operators own it. The people doing the changeover know where the time hides, and they are the ones who have to keep the new method alive.

Benefits of SMED

Faster changeovers pay off in several directions at once. They return capacity, because changeover time is downtime and cutting it lifts the availability side of OEE directly. They make small-batch production economical, which lowers inventory and work in process and shortens lead times. They make the plant more flexible and responsive, able to switch products to match demand instead of building to a forecast. And they tend to improve quality, because eliminating post-restart adjustments removes a common source of scrap at the start of a run. Implementations commonly cut changeover time by 50 to 80 percent within months, and the best programs, following Shingo's documented results, average around 94 percent.

Where SMED has the biggest impact

SMED matters most wherever changeovers are frequent and dominate downtime. In changeover-heavy operations such as food and beverage, pharmaceutical and consumer packaging, and multi-product discrete manufacturing, changeover can account for a large share of all downtime, often in the range of 20 to 40 percent. In those settings, attacking changeover is frequently the fastest route to more capacity without buying another machine. Plants that run long campaigns of a single product see less from SMED, though even there it buys flexibility.

Common obstacles

SMED is conceptually simple, but it runs into predictable resistance. Experienced operators who have run a changeover the same way for years may not see the point of a new method, so involving them early and crediting their knowledge matters. There is usually some upfront cost in tooling, clamps, and staging equipment, which has to be weighed against the recovered capacity. Training takes time. And the most common failure is treating SMED as a one-time event rather than a standard that has to be held: without follow-up, changeover times quietly creep back up.

SMED and the rest of lean

SMED does not stand alone. It is the enabler of just-in-time and pull production, because neither works without fast changeovers. It pairs naturally with 5S, since a place for every tool and a clean staging area cut search time out of every changeover. It feeds on Kaizen, the continuous-improvement habit that keeps changeovers shrinking. And it sits squarely inside the OEE framework as the main lever on the setup-and-adjustments loss, which is why it is so often the first improvement a plant reaches for after it starts measuring the Six Big Losses.

Measuring SMED

The headline metric is simple: changeover time, measured the strict way, from last good part to first good part at full speed. Beyond the average, watch the spread, because a changeover that swings from eight minutes to thirty is not really under control, and track the number of changeovers, since the whole point is to make more, smaller runs viable. Roll all of it up into the availability component of OEE, where the recovered time shows up as real gained capacity. Tracking changeover by product and by line is what tells you where the next event should go.

From a faster changeover to a faster plant

A SMED event produces a fast changeover on one machine on one day. Keeping it fast across every line, every product, and every shift is a different problem. Standards slip, a staging cart goes missing, a new product gets a sloppy first changeover, and the times drift back up without anyone deciding to let them. The gains are real, but they erode quietly unless someone is watching.

This is where SteelTree fits. It tracks changeover performance across your lines from the live data, surfaces the products and shifts where setups are running long, flags when a line is drifting off its standard, and routes the fix to the right person before the loss compounds. And because it captures what actually brought a changeover down, the next line starts from what already worked instead of from scratch. SMED makes the changeover fast; SteelTree keeps it that way.

See how SteelTree turns operational data into decisions →