A changeover is the time from the last piece of the completed run to the first good piece of the new run. SMED shortens it by reducing the work done while the machine is stopped. The steps, the measurement and the most common mistake.
5 November 2026 · 5 min read
Sample company · demo dataSMED steps
The Lean Enterprise Institute lexicon defines a changeover as a machine or line switching from one product to another and measures changeover time from the last piece of the completed run to the first good piece of the new run. The definition matters: what counts is not the time the machine is stopped but the time it is not making good parts. The first parts scrapped after adjustment are part of the changeover too. SMED is the method for shortening that time systematically, and the lexicon's setup reduction entry summarises it in six steps.
Why short changeovers?
A long changeover means large batches: if a changeover is expensive, the plant makes a lot of the same product, stock grows and the response to customer orders slows. When the changeover gets shorter, working in small batches becomes possible; stock falls, lead time shrinks and the line can switch between products more often. The TPM page of the US Environmental Protection Agency counts setup and adjustment losses as one of the six big equipment losses; SMED is the direct remedy for that loss.
Internal work, external work
The basic distinction of SMED is this: work that must be done while the machine is stopped is internal; work that can be done while the machine is running is external. Fetching the new die from the store, preparing fixtures, preheating and finding the next product's paperwork are external work, yet in most plants they are done after the machine has stopped. The method's first big gain is moving that work to while the machine is running.
The six steps
Measure: watch one changeover from start to finish; write down every step and its time. A video recording ends the argument.
Separate: mark every step as internal or external. Change nothing at this stage; only classify.
Convert: move work that looks internal but can in fact be done while the machine runs to external. In most changeovers the biggest gain is here, and it costs almost nothing.
Shorten internal work: shorten the remaining internal steps with quick connections, standard heights, pre-set fixtures and parallel work.
Shorten external work: organise the preparation too; external work does not stop the machine but it takes the operator's time.
Standardise: write the new sequence, teach it to the operators and measure again at the next changeover.
Which changeover, measured how many times?
Not every changeover is the same; the most frequent or the longest one is chosen first. A single measurement misleads: if at least two or three changeovers are observed on different shifts and with different operators, it becomes clear which steps are done every time and which depend on the person. If the measurement is done by the team that performs the changeover rather than by an engineer, the steps are named in the floor's own words and the later standard is adopted more easily. The target is set in the same unit: not total changeover time but the internal time the machine is stopped. Even if external work does not shrink, a drop in internal time turns directly into capacity.
Parallel work
Two tasks that two people could do at the same time take twice as long when done one after the other. If one person connects the hydraulics while another connects the sensor cables, only the longer of the two enters the downtime. Parallel work requires the second person to be ready at exactly that moment, which means the changeover is called by a schedule and a signal.
The most common mistake
The most common mistake is skipping the third step and buying quick connections and expensive fixtures straight away. Yet the work that can be moved external usually makes up the greater part of the time and moves at close to zero cost. The second mistake is not measuring the new sequence: if the standard is written but the next changeover is not measured, the gain is an assumption. The third mistake is leaving the operator out of the study; the person who does the changeover every day knows better than the engineer which step is unnecessary.
Sample company · demo data
Stopwatch step timing
How it is done in Gemy
An SMED study is opened in the time study section; steps are timed with a stopwatch and listed. The company switches the feature on.
Three phases work on the same step list: in Separate each step is marked internal or external; in Convert internal steps that could be made external are marked; in Shorten each step gets a target (keep as is, convert to external, shorten or eliminate), a target time and a parallel flag.
Current downtime, target downtime and time gain are calculated; of parallel steps carrying the same sequence number only the longest enters the total. Under every figure the calculation is written line by line.
An action can be opened directly from a step to be shortened; the study can be linked to a Kobetsu project and a gain record. Before and after measurements sit side by side.
Gemy does not take times from the machine automatically; the measurement is done by hand with a stopwatch.
In short
SMED measures changeover time from the last piece of the completed run to the first good piece of the new run and shortens it in six steps: measure, separate, convert, shorten internal work, shorten external work, standardise. The biggest and cheapest gain is moving work done while the machine is stopped to while it is running. The new sequence does not count as a gain until it is measured again at the next changeover; the operator who performs the changeover is at the centre of the study.
Takt time is the customer's pace; Yamazumi stacks each station's load against that pace. Calculating takt, reading the chart, the balancing sequence and how to explain balancing to the operator.
Autonomous maintenance means the person who knows the machine best, the operator, takes over its daily care. The content of the first three steps, the abnormality tag, the CIL standard and the division of work between operator and maintenance.
Kamishibai checks with short cards, at regular intervals, whether a standard is being followed on the floor. The structure of a card, the layers, good and bad question examples and the path of a red card.
7 November 2026 · 5 min
Cookies
Essential cookies are needed for the site to work. Analytics cookies load only if you allow them; declining is as easy as accepting. You can change your choice at any time via the footer link.