SMED: How to Reduce Changeover Time

Steps to cut a changeover

1. 1

   Define where the changeover starts and ends

   Use the standard definition: from the last good part of the old product to the first good part of the new one, at production rate. This matters because it is the only definition that captures what actually costs you — the ramp-up, the first-off inspection, the two scrap parts while settings settle. Teams that measure from 'machine stops' to 'machine starts' routinely report a changeover half the length of the real one, and then cannot explain why the schedule still slips.

2. 2

   Record one complete changeover on video

   Film it end to end, including the walking and the waiting. A changeover is the worst possible candidate for stopwatch study: it is irregular, several people move at once, and the observer cannot follow all of them. Video lets you replay each participant separately, catch the trip to the tool crib nobody mentions, and time steps that were over before you could react. It also settles arguments later, when someone insists a step takes two minutes and the recording shows six.

3. 3

   List every step with its time and its owner

   Break the recording into individual steps — fetch tool, remove clamp, lift die, clean, position, align, bolt, adjust, trial run, inspect. Record who does each one and how long it takes. Aim for steps small enough to be moved or reassigned individually; 'change the die' is one line and useless, while the twenty steps it consists of are where the time actually is.

4. 4

   Classify each step as internal or external

   Internal means it can only be done while the machine is stopped. External means it could be done while the machine is still running the previous product. Be strict: 'we always do it during the stop' is not the same as 'it must be done during the stop'. This single classification is the heart of SMED, and the first pass through it usually shows a surprising share of the stopped window spent on work that never required the stop — fetching, staging, preparing, looking for things.

5. 5

   Move existing external work out of the stoppage

   Anything already classified as external and currently done inside the stop moves out immediately: pre-stage the die on a cart, kit the tools and fasteners, print the paperwork, fetch the first-off gauge. This costs nothing but organisation and is the fastest gain available. How much it saves depends entirely on how much of your stopped window turned out to be external — which is why step 4 is measured rather than assumed.

6. 6

   Convert internal steps into external ones

   Now attack the steps that genuinely required the stop and redesign them so they no longer do. Standard approaches: pre-heat dies or moulds so warm-up happens off-line; use a second die holder so the next set is assembled while the current one runs; pre-set tools in a fixture outside the machine; standardise die heights and clamp positions so a shim stack does not have to be rebuilt each time. Each conversion moves minutes from the stopped window into running time.

7. 7

   Eliminate adjustment — it is not setup, it is trial and error

   Adjustment is usually the largest remaining block and the most misunderstood, because it looks like skilled work. It is repeated approximation: set, trial, measure, correct, repeat. Replace it with settings that cannot be wrong — fixed stops, locating pins, numeric scales with recorded values per product, dedicated spacers instead of measured ones. A position that is dialled in by feel takes minutes and varies by operator; a position that drops onto a pin takes seconds and does not.

8. 8

   Streamline what is left

   Speed up the steps that must stay internal: functional clamps and quarter-turn fasteners instead of bolts with long threads, cam locks instead of screws, standard fastener sizes so one tool does the job, and tools stored at the machine rather than in a crib. The rule of thumb worth remembering is that a bolt only fastens on its last turn — every earlier turn is waste that a quick-release mechanism removes.

9. 9

   Run steps in parallel where safe

   Many changeovers are serial only because one person does them. Two people working on opposite sides of a machine can halve elapsed time without any equipment change — provided the sequence is written down, roles are assigned, and the safety interlocks and lockout procedure account for both. This is a genuine gain in line availability, not in labour hours: the same total work happens in less stopped time. Never trade a safety step for parallelism.

10. 10

    Write the new standard and verify it holds

    Document the new sequence with roles, timings and the external preparation checklist, then run several changeovers against it and measure. Expect the first one to be slower than the plan. What matters is the trend over the next five, and whether the external preparation actually happens before the stop — that is the part which quietly decays first. Re-record a changeover a month later; if the time has crept back, the preparation checklist is usually the reason.

Before committing to the conversion work it is worth knowing what the minutes are worth in capacity — put a number on the changeover time you would win back.