SMED Changeover Analysis

What this tool does

CM8-214 is a working SMED study. You time a changeover task by task, classify each task as internal (the machine is stopped) or external (it could happen while the machine runs), record how each task could move or shrink, and the tool projects the changeover you would have if every improvement landed. It then tracks each improvement to a new standard and prints a report you can put in front of a production meeting.

Everything runs inside this single file. There is no account, no upload and no network request of any kind — your cycle times and machine names stay on the computer you are using.

What SMED is

SMED — single-minute exchange of die — is the changeover-reduction method developed by Shigeo Shingo from his work in Japanese manufacturing, most famously on press changeovers. "Single minute" does not mean one minute: it means a single-digit number of minutes — under ten. Shingo's changeovers routinely fell from hours to minutes, and the striking thing about how they fell is that most of the reduction came before anyone modified a machine.

The method's insight is that a changeover is not one lump of downtime but a sequence of tasks, and most of those tasks do not actually require the machine to be stopped. They happen during the stop only because nobody has separated them out. Do that separation honestly and the downtime collapses.

Internal and external — the distinction that does the work

Every task in a changeover is one of two things:

• Internal — it genuinely requires the machine to be stopped. Removing a tool from the press is internal: you cannot do it mid-stroke.
• External — it can be done while the machine is still running the previous job, or after it has restarted on the next one. Fetching the tool from the store is external: the press does not care where you are standing.

This one distinction carries almost the whole method. The changeover time that matters — the time the machine earns nothing — is the internal time only. Every minute you reclassify from internal to external comes off the changeover without making anyone work faster. The register asks the question twice for each task: what is it now, and what could it be? The gap between those two answers is where the tool flags "Move out" — the classic SMED finding, and there are usually more of them than anyone expects.

Step 1 — observe and time

Stand at the machine with a stopwatch through one complete changeover, from the last good part of the old job to the first good part of the new one. Write down every task in the order it happens, with its minutes — including the walking, the searching, the waiting and the second trip to the store. Those are the tasks people leave out because they are embarrassing, and they are exactly the ones the method feeds on.

Enter one row per task, in order. Resist rounding: 8 minutes of paperwork is a different fact from "about 5". Use the notes field for what the stopwatch could not capture — why the second trip happened, what the operator was waiting for.

Step 2 — separate external from internal

Go down the list and mark each task's current classification. Be strict: a task is only internal if the machine physically cannot run while it is done. In a first study it is common to find that half the elapsed time was external work being done during the stop — paperwork, fetching, staging, booking — for no reason other than habit.

Then answer the second question for every internal task: could it be external? "Yes" means the whole task moves outside the stop. "Partly" means some of it can — pre-setting a gauge from the job card is external, the final first-off check is not. "No" should be rare and defended: it is reserved for work that truly needs the machine stationary.

Step 3 — convert internal to external

Converting is mostly preparation and staging, and it is the cheapest improvement in manufacturing: print the pack during the previous run; stage the tooling on a trolley at the machine; deliver the next material to a marked square before the run ends; bring the checker to the machine instead of walking the part to the checker. None of that needs capital. It needs someone to own the preparation and a trigger — usually "when the run has thirty minutes left" — for it to start.

Write the mechanism in the improvement idea field. The tool refuses a "yes, it can move" without an idea recorded, because "we should do this outside the stop" without a how is a wish, not a plan.

Step 4 — streamline what remains

Only now, with the genuinely internal work isolated, do you make tasks faster. Enter a realistic streamlined time for each — what the task could take, not what you hope. The classic moves: replace threaded fasteners with quick-release clamps (a bolt's last turn does the clamping; the other nine turns are waste); standardise tool heights and add locating pins so a tool seats without measuring; pre-set anything that can be pre-set; put two people on tasks that parallel cleanly. External tasks are worth streamlining too — they cost labour even when they no longer cost downtime — but the internal ones come first.

The biggest win is usually organisational, not mechanical

Look at where the minutes go in the sample study: paperwork, fetching tooling, walking to quality, booking the job, staging material. None of that is machining, and all of it moves outside the stop with nothing more than preparation and a trolley. The mechanical improvements — quick clamps, standard heights — matter, but in a first study they are typically the smaller half of the saving. This is good news: it means the first big cut in changeover time costs almost nothing and needs no engineering approval, which is why SMED is usually the first lean tool worth running on a bottleneck machine.

Film the changeover

If the team will tolerate it, film one changeover and watch it together with the task list on the table. Film the work, not the person, say so beforehand, and delete the footage after the review. Video settles arguments a stopwatch cannot: how long the search actually took, how many trips there really were, where two people stood watching one. The person doing the changeover should be in the review — they know which minutes were absurd better than anyone, and the improvements they propose are the ones that stick.

The economics — why shorter changeovers matter

The obvious benefit is recovered capacity on the machine. The larger one is batch size. Long changeovers force long runs, because the changeover cost has to be spread over enough parts to bear it; long runs mean weeks of stock, money sitting on shelves, and slow response when the schedule changes. Cut the changeover and smaller batches become economic, which cuts inventory and lead time together. Be honest about the limit of this: a shorter changeover permits smaller batches — it does not deliver them. If the planning system keeps launching the same big runs, you will simply have a faster changeover and the same stock.

Standardising the new method

That is what the status field is for, and it is the whole point of the last step. An improvement that is observed is a fact; planned is an intention; trialled means it has been run at least once; standard means it is written into the changeover sequence, the trolley layout and the training — the new normal that the next study starts from. A study where every idea stays at "planned" has produced a document, not a shorter changeover. The implementation plan table lists everything still short of standard, with its owner and the minutes at stake, in order of size — work it from the top.

The formulas

The current changeover tile sums every observed task in the latest study; the internal figure beside it counts only the tasks done while the machine is stopped, which is the downtime that matters. The projected figure assumes every "move it out" moves and every streamlined time is achieved — it is a target, and the before-and-after chart says so.

FAQ

Which machine should we study first? The bottleneck — the machine the whole flow waits for. An hour saved there is an hour of output; an hour saved on a machine with spare capacity is just a tidier changeover.

How many tasks should a study have? Ten to twenty-five is typical. If you have five, the tasks are too coarse to act on — "set the tool, 30 minutes" hides the six decisions inside it. If you have sixty, combine steps nobody will manage separately.

What if two people work the changeover at once? Record the tasks each person does with their own times, in overall sequence order, and note the parallelism in the task notes. The projected total will overstate the elapsed time when tasks genuinely overlap — read it as work content, and time the real elapsed changeover at the trial.

Our changeover is already under an hour. Is this worth doing? If the machine matters, yes. The method scales down: a 40-minute changeover with 15 minutes of fetching and paperwork inside the stop has the same disease as a four-hour one.

Can one register hold several machines? Yes — each machine-and-date combination is a separate study, and the tiles, charts and summary tables always show the latest one. Export the register before starting a new study if you want the old analysis preserved as its own file.

Saving your work

Tasks, settings and the report header are written to this browser's local storage as you type, and the toolbar shows the time of the last save. That storage belongs to one browser on one computer: another browser, a private window, a second machine or a clean-up tool that clears site data will not have it.

Treat Export .json as the real save — one file containing everything, which Import .json restores anywhere. Export CSV gives you the task list for spreadsheet work. Reset asks twice, then erases everything this tool has stored. There is no undo.

Accuracy & disclaimer

The arithmetic here is simple and the tool does it faithfully. The judgement underneath it is yours: whether a task is genuinely internal, whether a streamlined time is realistic, and whether the observed changeover was a typical one. One study is a snapshot — observe two or three changeovers before locking the new standard, because operators, tool pairs and days differ.

The projected changeover assumes every improvement is implemented and works; treat it as a target to be proved at a trial, not a result. Physical changes to clamping, tooling or guarding must be assessed and approved by whoever is competent for that machine and your local requirements. This is an analysis and record-keeping aid, not engineering advice.