How to implement 5S in your factory, step by step
Rolling out 5S is easy. Having it still standing in six months isn't. That's where you find out whether you built a system or ran a cleaning campaign.
Rolling out 5S takes five stages: sort, set in order, shine, standardize and sustain.
Enter your email and you will get an alert when something new is published. Unsubscribe anytime.
But the real work isn't in the first four. It's in the fifth. It's the only one that still costs effort when nobody is watching any more.
The technique was born at Toyota in the sixties and arrived here with a reputation as a cleaning method. That misunderstanding is the main reason it fails, and it's been doing it for sixty years.
Where you start
One pilot area, not the whole plant. A bounded area, with an owner who has a name, and if you can manage it, somewhere visible from the rest of the floor. Making it work in one place convinces people more than any presentation you can put together.
Take the before photo. It looks like a trivial thing and it isn't. In three months nobody will remember what that place looked like, and that photo is what holds the team's morale up when it sags.
Sort with a criterion, not by eye. Anything not used in the last month goes out of the area. If there's an argument — and there will be — it goes to a quarantine zone with a date: thirty days unclaimed and it's gone.
Set things in order at hand height. What's used every hour, within reach. What's used weekly, further away. And mark the places, so the empty gap sings out at you.
Clean in order to inspect. This is the change of mind that decides everything: you don't clean to make it pretty. You clean to find the leak, the play, the bolt that's vibrating loose.
Standardize with a photo, not a manual. One image of the correct state posted at the station beats a six-page procedure that not one human being has ever opened.
The three mistakes that sink it
| Mistake | How it looks | What to do |
|---|---|---|
| Treating it as cleaning | it happens before the customer visit | set the goal on detecting, not presenting |
| Imposing it from above | designed by someone who isn't at the station | let the person who works there arrange it |
| No auditing | three months on, everything's back | short, regular audit with the area owner |
How it gets sustained
There's no trick here and there isn't going to be one: short audit, frequent, with a consequence. Ten minutes, a short list, same time every week, and something happening with the result. It gets discussed, compared, recognized.
And there's one condition almost everyone skips: the time for 5S has to sit inside the working hours. If it happens "when we can", it doesn't happen. Ever. Fifteen assigned minutes at the end of the shift are worth more than a full day twice a year.
The bridge to 2026
5S organizes physical space so the anomaly is visible. What I've spent twenty-five years deploying does the same with the process: makes the deviation visible the moment it occurs.
And there's a dependency between the two you'd do well to understand before spending a euro on technology.
A plant without 5S is not ready for AI. And I don't say that out of philosophy. If material has no fixed place, if the light changes from one day to the next, if every shift arranges things its own way, the images you're feeding the system are noise. Full stop.
The computer-vision projects I've watched run aground didn't fail because of the model. They failed because the scene changed every day and nobody had stopped to look.
5S doesn't come before AI because it's humbler. It comes before because it's what makes the data worth anything.