Mistake-Proofing vs. Inspection: What’s the Difference?

Mistake-proofing — poka-yoke, in the term Shigeo Shingo gave it at Toyota — designs a process so that an error either cannot occur or is caught the instant it does. Inspection examines output after the fact to find the defects already made. The distinction is the difference between prevention and detection, and it carries a cost structure: mistake-proofing is paid for once, while inspection is paid for on every unit, forever.

How they differ

Inspection is detection downstream. It never adds quality; it sorts whatever quality already exists, and it sorts imperfectly — sampling lets defects through by design, and even 100 percent human inspection fatigues, drifts, and misses. Mistake-proofing moves the control into the work itself: a connector shaped to fit only one way, a fixture that will not close over a missing part, a form that will not submit with an empty field. The best devices make the error physically impossible; the next best make it immediately visible at the point of work, while the unit is still in hand and the fix costs seconds.

The two are not enemies. Where consequences are severe or the devices are imperfect, inspection remains a sensible backstop. The direction of travel is what matters: every recurring find at final inspection is an invitation to design the error out upstream. Shingo’s deeper point was about timing: catch the error — the cause — rather than the defect, its consequence, and the inspection debate largely dissolves.

An illustration: the missing screw

An electronics assembler keeps catching units at end-of-line test with one fastener absent — findable, but only after full assembly, and rework means partial teardown. The fix costs almost nothing: technicians draw screws from a pre-counted tray, one cavity per screw, one tray per unit. An empty tray means a complete unit; a cavity still holding a screw stops the unit before it leaves the bench. The defect no longer survives to the end of the line, and the end-of-line check quietly stops finding it.

• Prefer designs that make the error impossible; settle for signals that make it obvious
• Put detection at the point of work, not at the end of the line
• Read every recurring inspection find as an upstream design gap
• Favor cheap and boring devices — trays, templates, one-way fits — over clever ones
• Keep inspection where the stakes are high; retire it only after the device proves itself

Inspection pays rent on a problem; mistake-proofing ends the lease.

Mistake-proofing enters the curriculum at Yellow Belt with the core Lean ideas and becomes a working tool in the Improve and Control phases at Green Belt, where a fix is not finished until something stops it from regressing. Both programs end in proctored exams with a 70 percent passing score and one free retake.