Machinery Wiki

Changeover and SMED

SMED (single-minute exchange of die) is the discipline of turning changeover from an internal, line-stopped activity into external preparation plus a short, repeatable internal routine. This page covers the eight-step method, what to measure, and why most changeover programmes stall.

By Global Machinery Hub · Published November 2, 2026 · Updated November 2, 2026 · 6 min read
Changeover & SMED

The core idea of this page: changeover time is not fixed by the machine — it is fixed by how much of the work has to happen while the line is stopped. SMED attacks that split. Everything that can be prepared, pre-heated, pre-positioned or pre-checked while the line still runs is external; what remains is internal, and internal work is what you shrink.

Why changeover is the highest-leverage loss on a packaging line

Breakdowns are unpredictable; changeovers are scheduled, repeatable and therefore improvable. A line that loses ninety minutes per changeover four times a day loses six hours a day — often more capacity than all its breakdowns combined. Because the activity is repeatable, improvements compound: every minute removed is removed every time.

The eight-step method

1. Observe and record

Film the changeover and log every action with a start and stop time, including walking and waiting. Film is uncomfortable and therefore effective: it exposes the searching, the second trip to the tool room and the discussion about which die goes in first.

2. Separate internal from external

Mark every logged action as internal (line stopped) or external (line running). The first pass typically finds a third of internal time that could be external — tools, documentation, next format's parts, cleaning supplies.

3. Convert internal to external

Build a changeover cart with the next format's tooling, gauges and spares. Pre-heat sealing jaws. Pre-stage film and labels. Pre-print the settings sheet. This step alone often halves changeover time without touching the machine.

4. Streamline the internal routine

Then attack what is left: quick-release fasteners instead of bolts, captive screws, colour-coded and keyed parts that cannot go in the wrong place, numbered steps, and a two-person routine split so nobody waits.

5. Eliminate adjustment

Adjustment is the enemy of repeatability. Wherever a setting is found by trial, replace it with a fixed stop, a scale, a pin position or a stored recipe. Every "turn it until it runs" step is a hidden minute and a source of variation.

6. Standardise the routine

Write the sequence down, with photographs and torque values. The standard is what makes the time predictable — and predictable time is what lets planning promise a real changeover, not an average.

7. Train and practise

Run the routine as a drill, timed, with the same crew. Improvement from practise alone is real: crews that rehearse cut minutes that no capital spend would have found.

8. Sustain with measurement

Log every changeover time on a visible chart by format and crew. Investigate the outliers; do not average them away. The chart is the control, not the celebration.

What the target should be

The name "single minute" refers to single-digit minutes, and it is achievable on well-designed equipment when the routine is genuinely external-heavy. For most packaging lines the practical sequence is: baseline, remove external work, then halve, then keep going. Toyota's own trajectory is a useful calibration — hours to fifteen minutes, then three, then under three — and it took decades of iteration rather than one project.

Building the baseline and the business case

Before changing anything, record twenty changeovers. For each, capture total time, format, crew, shift and the three numbers above. The distribution matters more than the average: if the best changeover is sixty minutes and the worst is two hours, the routine is not standardised and the gap between best and worst is capacity you already own.

Then convert minutes into money honestly. Multiply the minutes saved per changeover by the number of changeovers per year, then by the line's contribution margin per minute at the bottleneck — not by a fully loaded overhead rate. If the line is not the constraint, extra minutes may only build inventory, and the business case should say so before anyone buys hardware.

When quick-change hardware pays

Quick-release clamps, plug-in format sets, servo-driven adjustments with stored recipes and magnetic tooling all shorten internal work, but they are only worth their price once the routine, the external preparation and the adjustment elimination are done. Buying hardware first usually produces a machine that changes format in fifteen minutes while the crew still spends forty-five minutes looking for the gauges.

What to measure

Measure from last good pack of the previous format to first good pack of the next, not from "machine stopped" to "machine started". That end-to-end definition captures startup scrap and first-off approval, which are often the largest hidden parts of a changeover.

Record three numbers per changeover: total time, number of adjustments made after restart, and units scrapped before first good pack. The second and third tell you whether the routine is standardised or being improvised.

Failure modes that stall programmes

Three patterns recur. First, treating SMED as a one-off workshop: the cart gets built, then borrowed from, then empty. Second, pushing internal work into external time on paper while the crew still does it with the line stopped. Third, buying quick-change hardware without changing the routine, which shortens the bolt-turning but leaves the searching and the adjustment untouched.

Verification checklist

Confirm: the changeover definition runs to first good pack; the external checklist is complete before the line stops; the internal routine has a written sequence with photographs; adjustments after restart are counted and trending down; and scrap before first good pack is tracked. If all five hold, the changeover is under control.

Source: Single-minute exchange of die — Wikipedia

"Over time, Toyota decreased changeover times from hours to fifteen minutes by the 1960s, three minutes by the 1970s, and ultimately just 180 seconds by the 1990s."

Frequently asked questions

Does SMED require new hardware?

Not first. The largest early gains come from moving work external and standardising the routine; quick-release hardware is most valuable after the routine is stable, when it removes the remaining internal minutes.

How many people should do the changeover?

Whatever the routine defines, and the same number every time. Adding people ad hoc hides the real time and makes improvement impossible to measure.

Is changeover time the same as setup time?

Setup usually means the machine-side work; changeover is the end-to-end activity including cleaning, documentation and first-off approval. Plan with the end-to-end number.

Related reading: changeover optimization guide, OEE, doypack formats, and the form-fill-seal category.

Related in Packaging Machinery