Poka-Yoke: Mistake-Proofing in Practice

Poka-yoke, usually translated as mistake-proofing, is the practice of designing work so that an error cannot turn into a defect, or is caught the moment it occurs. The term comes from Shigeo Shingo and the Toyota Production System, where it replaced the idea that quality is a matter of telling people to be more careful. People will always make occasional slips; a good poka-yoke makes the slip harmless. This guide covers what a poka-yoke actually is, the difference between control and warning types, the three detection methods, real examples, and how it fits alongside inspection and analysis tools like FMEA.

Error is not the same as defect

The founding insight of mistake-proofing is the gap between an error and a defect. An error is the slip: a part fitted backwards, a step skipped, a wrong component picked from a bin. A defect is what the customer receives when that error is not caught. Between the two there is always a window, and poka-yoke lives in that window. The goal is not to stop people making mistakes, which is impossible, but to stop mistakes reaching the next process or the customer.

This reframes quality from an attitude problem into a design problem. If an operator can load a fixture the wrong way round, that is not primarily a training failure; it is a fixture that permits the wrong way. Shingo called the ideal source inspection: checking for the conditions that cause a defect, at the point they occur, rather than inspecting finished parts and sorting good from bad. A poka-yoke is how source inspection gets built into the work itself.

Control vs warning

Poka-yoke devices fall into two families by how strongly they act.

• Control (prevention): the strongest form. The device physically prevents the error or halts the process until the condition is right. A connector that only mates one way, a fixture that will not clamp unless the part is seated, a machine that will not cycle until a light curtain confirms the guard is clear. Nothing depends on the operator noticing.
• Warning (detection): the device signals that something is wrong (a light, a buzzer, a screen message) but the process can continue. It is weaker because it still relies on a human to see the signal and act. Use it where a full control device is impractical, and make the signal impossible to ignore.

Given a choice, always reach for control over warning. A warning that fires forty times a shift becomes background noise; a control device that will not let the wrong thing happen never fatigues.

The three detection methods

Independently of control versus warning, Shingo classified poka-yoke by how the error is detected. Most real devices are one of these three:

• 1. Contact method: uses physical shape, size or other properties to detect contact or its absence. Asymmetric locating pins, a slot that only accepts a correctly oriented part, a probe that must touch the workpiece before the cycle starts.
• 2. Fixed-value (constant number) method: checks that a set number of actions or parts has occurred. A dispenser that releases exactly the right count of screws, so a leftover screw means a step was missed; a kit tray with one recess per component.
• 3. Motion-step (sequence) method: checks that steps happen in the correct order. The machine will not advance to step three until sensors confirm steps one and two, so a skipped operation stops the line rather than shipping.

Examples on the floor

Good poka-yoke is usually cheap and obvious in hindsight. A few patterns recur across plants:

• A fixture with an asymmetric locating pin so an asymmetric casting cannot be loaded rotated 180 degrees.
• A light curtain or interlock that stops a press cycling while a hand is in the die area, which is a safety poka-yoke and a quality one.
• A pick-to-light or kitting tray that lights the next bin and will not confirm assembly until the correct part is drawn.
• A checkweigher or a vision inspection station that rejects any pack outside a set weight or missing a label before it leaves the cell.
• A torque tool that will not send its done signal until the correct number of fasteners has hit target torque (a fixed-value method).

What it is worth: a small example

Poka-yoke has no headline formula, but its payback is easy to size because it works on the scrap rate and cost of poor quality. Suppose a cell runs 4,000 parts a shift and one recurring assembly error escapes at 1.5%, so about 60 defective parts a shift. If each escape costs 8 in scrap, rework and handling, that is roughly 480 a shift, or over 100,000 a year on two shifts. A locating-pin fixture that makes the error physically impossible costs a few hundred to build. Even if it only removes that single failure mode, the return is not a close call. Sizing the loss first, in your own numbers, is what turns mistake-proofing from a nice idea into a funded project.

Poka-yoke, FMEA and inspection

Mistake-proofing does not stand alone. It pairs naturally with the tools around it. An FMEA ranks where failures could happen and how risky each is; a high detection or severity score is a signal that a poka-yoke is worth building at that step. Root cause analysis on a defect that already escaped often ends with a poka-yoke as the permanent countermeasure, so the same fault cannot recur. And where a defect maps onto one of the six big losses, mistake-proofing is usually the cheapest way to attack the quality losses at source.

It also changes the role of inspection rather than removing it. Final inspection finds defects after the cost is sunk; source inspection built into a poka-yoke catches the condition that causes them. You will still verify and audit, but the aim is to need far less sorting because fewer bad parts are ever made.

Where to start

Start with the defect that hurts most, not the easiest device to build. Pick a recurring, well-understood error, work out where in the sequence it happens, and design the simplest device that either prevents it or catches it at that point. Prefer control over warning, keep it cheap, and test that it actually blocks the failure mode rather than just looking clever. Then measure: track first pass yield and scrap on that operation before and after, so the effect is visible in the data and the next project earns its budget.

Judge mistake-proofing by whether defects at the source actually fall, which needs honest defect and stop data rather than memory. The partner we recommend, Fabrico, captures stops and quality events directly from the line and shows the true cause on video, so you can see whether a new device removed the failure mode or just moved it. Fabrico is a partner we recommend; the tools here are free regardless.

FAQ

Related: FMEA · root cause analysis · the 8 wastes of lean · the six big losses · scrap rate calculator