SMED: Reduce Changeover Time and Batch Sizes
Long changeovers lead to large batches, and large batches lead to high inventory and long lead times. SMED reduces changeover time systematically, usually without any up-front investment. This lesson shows the four steps and uses three examples to calculate how a shorter changeover affects effective cycle time, batch size, EPEI and inventory.
What SMED is
SMED stands for Single Minute Exchange of Die, meaning a die change in single-digit minutes. The method goes back to Shigeo Shingo, who developed it from the 1950s onward, at Toyota among others. The goal is not to work more frantically, but a changeover sequence in which the machine stops only for the steps that truly require it to be stopped.
Changeover time is the time from the last good part of the old batch to the first good part of the new batch at normal speed. Start-up, trial parts and readjustment are included, not just the physical swap. In the data box it appears as C/O (changeover) in minutes.
Internal and external setup
Internal setup is everything that can only be done with the machine stopped: removing and installing the die, clamping, adjusting, running a trial part. External setup runs in parallel with production: fetching and preheating the die, staging material, preparing clamps and gauges, loading programs, cleaning up after the changeover.
In many current-state changeovers the two are mixed: the machine sits idle while someone searches for the die or waits for the crane. This is exactly where SMED starts. For the value stream only internal setup counts, because that is the only time the machine loses capacity.
The four steps
- Analyze: video or time-study one real changeover from start to finish and record every action with its duration. Observe the actual sequence, not the one in the work instruction.
- Separate: mark each step as internal or external. Everything external is done before the stop or after start-up, for example with a changeover checklist and a setup cart.
- Convert: move internal setup to external, for example by presetting tools off the machine, preheating molds, or using intermediate clamping plates or zero-point clamping systems.
- Streamline: shorten the remaining internal steps, for example quick clamps instead of bolts, fixed stops instead of measuring and adjusting, standardized die heights, two people working in parallel. Make the new sequence the standard and practice it.
Changeover per piece in the data box
Effective cycle time accounts for changeovers through changeover time per piece: changeover time ÷ batch size. Uptime in the data box covers breakdowns only, so changeovers are not counted twice.
Example stamping in the steering bracket map: cycle time 1 s, uptime 85%, changeover 60 min, batch size 1,000. Changeover per piece = 3,600 s ÷ 1,000 = 3.6 s, effective cycle time = 1 ÷ 0.85 + 3.6 = 4.8 s. With SMED down to 10 min (the kaizen burst in the demo) it is 600 s ÷ 1,000 = 0.6 s and 1.8 s effective.
You can use the time gained in two ways: as capacity (same batch size, less changeover per piece) or for more frequent changeovers (smaller batches, same changeover per piece). With a 10 min changeover and 167 pcs per batch, changeover per piece stays at 3.6 s, but one batch now covers only about 0.2 of a day’s demand instead of a little more than one day (920 pcs). For the value stream the second option is usually the goal, because smaller batches mean less inventory.
Example pump assembly: changeover creates the bottleneck
In the pump assembly exercise, takt time is 450 min × 60 ÷ 300 pcs = 90 s. Machining has a 75 s cycle time, 85% uptime, 40 min changeover and a batch size of 150. Changeover per piece: 2,400 s ÷ 150 = 16 s, effective 75 ÷ 0.85 + 16 = 104.2 s. The process is above takt even though its plain cycle time is below it.
SMED to 10 min plus TPM to 95% uptime: at half the batch size (75 pcs), changeover costs 600 s ÷ 75 = 8 s per piece, effective 75 ÷ 0.95 + 8 = 86.9 s, so below takt. At the same time the batch inventory after the process is cut in half.
Changeover time, batch size and EPEI
When several variants share a machine, changeover time determines how often each variant can run. The time left after production is the changeover capacity. EPEI (Every Part Every Interval) is the sum of the changeover times for one cycle through all variants, divided by this changeover capacity per day.
Example milling in the template “Variant production”: 450 min working time × 80% uptime = 21,600 s. Load: 400 × 28 s + 250 × 24 s + 50 × 55 s = 19,950 s. Changeover capacity: 21,600 − 19,950 = 1,650 s, or 27.5 min per day. Changeover times of the three variants: 45 + 45 + 90 = 180 min = 10,800 s. EPEI = 10,800 ÷ 1,650 = 6.5 days.
With SMED on the special variant (90 → 30 min) the total drops to 120 min = 7,200 s, and EPEI to 7,200 ÷ 1,650 = 4.4 days. One batch per variant equals roughly daily demand × EPEI, and average batch inventory is about half a batch. For the standard variant (400 pcs per day) this means the batch drops from about 2,600 to about 1,750 pcs, and the average coverage of the batch inventory from about 3.3 to 2.2 days.
Planning SMED in the value stream
On the current state map, changeover problems show up in three places: a high C/O in the data box, large inventory after the process and an EPEI of several days or weeks. In the future state, SMED is usually the prerequisite for a supermarket with kanban and small batches; without a shorter changeover the supermarket gets large.
In the app, enter the action as a kaizen burst on the process, with owner and due date. In the future state, change changeover time and batch size in the data box and adjust the inventories to the new batch size. Effective cycle time, EPEI (via the product mix if there are several products), coverage and lead time are recalculated; the current/future comparison shows the effect.
Practice in the editor
Open “Variant production”, check the EPEI for milling in the product mix and reduce the changeover of the special variant (“Sonder”) there from 90 to 30 min. Compare EPEI and changeover capacity before and after.
Open the template in the editor Start blankFrequently asked questions
What does SMED mean?
Single Minute Exchange of Die: changeovers in single-digit minutes, so under ten minutes. The method comes from Shigeo Shingo. It describes a direction and a procedure, not a fixed limit for every machine.
What is the difference between internal and external setup?
Internal setup can only be done with the machine stopped (changing the die, adjusting, trial part). External setup happens while the machine is still running the old batch or already running the new one (staging the die, presetting, cleaning up). Only internal setup costs capacity.
Does changeover time belong in the uptime in the data box?
No. Uptime in the data box covers breakdowns only. Changeovers are accounted for in effective cycle time through changeover time ÷ batch size. In OEE, by contrast, changeovers count against availability.
Is SMED worth it if the machine is not a bottleneck?
Yes, if you use the time gained for smaller batches. Batch inventory, coverage and lead time then go down, even if capacity was already sufficient. If you only cut changeover time and keep the batch size, you gain spare capacity but no inventory reduction.
How do I measure changeover time correctly?
From the last good part of the old batch to the first good part of the new batch at normal speed. Start-up, trial parts and readjustment are included. Record several changeovers, because changeover times vary a lot.