Lean tools

Poka-yoke / mistake-proofing

In French: Poka-yoke (détrompeur)

A poka-yoke is a device that makes the error impossible, or that flags it the instant it occurs, before it becomes a defect.

What it is

Poka (inadvertent error) and yoke (avoid): the term, formalised by Shigeo Shingo, is rendered in English as "mistake-proofing" or "error-proofing". The principle rests on a simple observation, unflattering for the usual action plans: asking someone to be more careful does not lastingly reduce errors. Changing the device does.

The founding distinction is between the error and the defect. The error is the act — fitting the gasket the wrong way round. The defect is its consequence on the product. Between the two there is an instant, and that is where the poka-yoke acts: either it prevents the error, or it detects it before it spreads.

A good error-proofing device is cheap. The USB plug that only goes in one way, the jig that accepts the part only in the right orientation, the form that refuses an end date earlier than the start date: none requires a budget, all require having observed where the error occurs.

Prevent or detect

Preventing is always preferable: assembly is physically impossible the wrong way round, the machine does not start if the guard is open. The error does not happen, so there is nothing to catch.

Detecting is the fallback when preventing is not feasible: a sensor counts the screws and stops the conveyor if one is missing, a system alerts when an entered value falls outside the limits. The error happens but does not leave the station.

Shingo also described three ways of triggering the device: by contact (the physical shape does the sorting), by fixed value (a check that an expected number is reached — five screws, five welds), by sequence (the next step is blocked as long as the previous one is not validated). It is a useful design grid when one is looking for the idea and cannot find it.

What it is for, and where to place it

An error-proofing device serves to make a defect disappear rather than to catch it. The difference is economic before it is technical: an inspection catches part of the defects and costs on every part inspected, indefinitely; a poka-yoke costs once and lets nothing through. The calculation almost always tips the same way as soon as the defect recurs.

Where to place it follows from a single rule: as close as possible to the gesture, within the station itself. A device installed downstream no longer prevents anything, it sorts — and sorting means the defect has already been made. The right question is not "where could it be seen", it is "at what precise instant does the error become unrecoverable", and then you step back to just before.

It is also the reference action when an FMEA points to a high occurrence or a poor detection. On an FMEA line, an error-proofing device is about the only lever that moves those two scores lastingly — an instruction, a training session or an additional inspection move them on paper and rarely in production.

A badly screwed housing, figure by figure

Assembly station, 62,000 housings assembled per year, five screws to torque on each. Over the past year, 340 housings left the station with a screw missing or not tightened, i.e. about 5,500 per million.

The cost of the defect depends entirely on where it is caught, and the gap is brutal. At the station, putting a screw back takes 40 seconds. At final inspection, it has to be dismantled then reassembled: 12 minutes, i.e. eighteen times more. At the customer, between the finding, the investigation, the replacement, the transport and the file, it is counted in hours — several hundred times the cost of the same correction made at the station.

Of the 340 cases, 329 were caught at final inspection and 11 went to the customer. The 329 alone represent 66 hours of rework over the year. The 11 others cost far more and are not counted in hours only.

The device chosen detects nothing: it prevents. A counting screwdriver keeps the housing clamped as long as the five screws are not at torque; as long as the count is not reached, the part cannot leave the station. It is a fixed-value trigger in Shingo’s sense, and it requires neither supervision nor discipline — the operator who forgets a screw simply cannot move on to the next part.

Fourteen months later, not a single case. Final inspection on that specific point was removed, since it had nothing left to inspect. That last detail is the best test of a successful poka-yoke: if it had to be backed by an inspection to sleep soundly, then what was designed was not an error-proofing device but one more alarm.

The common mistakes

Calling an additional visual check a poka-yoke. Adding a box to tick on a sheet, or asking for a double check, is not error-proofing: it is shifting onto human attention what one did not want to handle through the device — and human attention is precisely the variable that failed. The test is simple: if the only thing preventing the error is that someone remembers something, it is not a poka-yoke.

Designing the device without the operator, then being surprised that it is bypassed. A device that slows the gesture or gets in the way of the grip ends up neutralised — shim removed, sensor masked — and the bypass is almost always rational from the point of view of the person doing it. It is not a discipline problem, it is a design flaw that would have been seen in thirty minutes at the station.

Error-proofing an error that never happens. The exercise is stimulating and weeks can be spent securing theoretical scenarios. A poka-yoke is designed from a defect that has been observed and counted, not from an imagined one — otherwise devices are added, hence complexity and breakdowns, for a risk that did not exist.

On the shop floor — Electronics assembly

A connector fitted the wrong way round came back a few times a month, despite a posted instruction and a final inspection. The fix was a two-millimetre notch in the holder, which made reverse assembly physically impossible. The defect disappeared, and the final inspection with it.

The other tools illustrated by this trade: electronics and aerospace.

Frequently asked questions

What is a poka-yoke, with an example?

It is a device that makes an error impossible, or flags it the instant it occurs. The most telling examples are everyday ones: the USB-C plug that connects either way up, the chamfered SIM card that only goes in one way, the oven door that cuts the heating element when opened, the form that refuses an end date earlier than the start date. None requires thinking about it — that is precisely the criterion.

What is the difference between a poka-yoke and quality inspection?

Quality inspection finds out after the fact that a defect exists, on all or part of production, and it costs on every part inspected. A poka-yoke acts before the defect exists, on every part, with no intervention. The test that separates them is simple: if the only thing preventing the error is that someone remembers something or looks carefully, it is not a poka-yoke.

What is the difference between poka-yoke and jidoka?

The two differ in their reach. The error-proofing device is local, attached to a precise gesture, and its ambition stops at that gesture. Jidoka is an organisational principle that concerns the whole line and commits to a chain of reaction: stop, call, treat the cause before restarting. A poka-yoke can exist in a workshop that does not practise jidoka; the reverse is rarer, because a workshop that allows itself to stop ends up wanting to remove the reasons for stopping.