Poka-yoke: prevention versus detection

Poka-yoke: prevention versus detection

Start with a specific error mechanism rather than a general request to be more careful. A keyed shape can prevent an incorrect orientation; a detection device can identify a missing part before it passes downstream. These are different levels of protection. Challenge the proposed control with known correct and incorrect cases, including plausible failure or bypass conditions. Define what happens when the check fails and who maintains the mechanism. A warning that people can overlook is not equivalent to physical prevention. Preserve required technical approval and verify the effect on quality before making the control part of the standard method.

Follow the method

1. Prevent error
2. Detect before escape
3. Challenge test
4. Maintain control

Show where prevention and detection interrupt the error path

Fictional connector assembly: the original connector can be inserted backward. Proposal A introduces a keyed geometry that should prevent that orientation. Proposal B uses a sensor to detect a backward insertion and block release. Engineer Sana must compare mechanisms rather than label both “mistake-proof” without testing.

Follow the method

1. Keyed geometry
2. Sensor + block
3. Correct parts only
4. Missing/wrong model
5. Fault response

A keyed part and a warning light solve different problems

Distinguish preventing a defined error from detecting it before escape, and build a challenge-test record that checks both the control and its failure response.

Fictional connector assembly: the original connector can be inserted backward. Proposal A introduces a keyed geometry that should prevent that orientation. Proposal B uses a sensor to detect a backward insertion and block release. Engineer Sana must compare mechanisms rather than label both “mistake-proof” without testing.

Role: Process engineer with operator, quality and design representatives.

Normal condition

The defined wrong orientation cannot reach accepted downstream output under the approved method.

The gap

The team has only tested correctly oriented parts and claims the wrong-orientation failure is eliminated.

• This is a desk comparison, not an instruction to bypass or modify production equipment.
• The control addresses a specific error mode; other wrong, missing or damaged parts require their own analysis.

1. Name the error precisely

   Sana defines the target as backward orientation of the specified connector at the insertion step. She separates it from missing connector, wrong model and damaged housing.

   Why: “Assembly errors” is too broad to support a testable prevention claim. A control may handle one mechanism while leaving others unchanged.

   Evidence: Target failure mode and point of occurrence are explicit.

2. Describe what each control does

   Proposal A is a prevention candidate if the approved geometry truly makes backward insertion impossible without damage or bypass. Proposal B is a detection/control candidate if sensing reliably blocks unsuitable release.

   Why: The distinction concerns the mechanism, not which proposal sounds more advanced. Detection can be useful, but it must have a dependable response.

   Evidence: A prevents the insertion mechanism; B detects and blocks escape under its validated conditions.

3. Challenge the failure modes safely

   Using approved challenge parts and test conditions, test correct orientation and the defined incorrect orientation. Include credible sensor fault, missing part or bypass scenarios where relevant to the control.

   Why: Passing normal parts tests availability, not necessarily protection. Deliberate challenge testing needs controlled conditions and traceable test objects.

   Evidence: Test matrix records expected and actual outcome for each defined case.

4. Verify the response to control failure

   Check that a failed or uncertain result leads to the required controlled state and named response, rather than a silent pass. Confirm reset/restart behavior through the approved method.

   Why: A detector that can fail unnoticed may provide false confidence. An audible warning alone may not prevent escape.

   Evidence: Fault/uncertain-state row includes containment, response owner and acceptance decision.

5. Maintain the control and its scope

   Document the validated failure modes, challenge frequency and change triggers. Recheck after relevant product, fixture, software or maintenance changes.

   Why: A successful initial test does not make the control permanent or universal. The operational standard must retain the challenge method and response.

   Evidence: Control owner, approved test record and scope limitations accompany release.

The sensor loses power

A fault test shows the sensor has no valid result after power loss, yet the proposed release signal remains permissive.

Do not approve the claimed detection control; the owner must correct and revalidate the fault response.

A normal-operation success cannot compensate for a known escape path in the tested fault condition.

Fault case recorded as failed, with affected claim and required retest.

Run a practice session

Materials

• Two design proposal cards and a blank challenge matrix
• Approved test-object placeholder; no live-machine challenge in class

1. Define the error · 4 minutes

   Which orientation and which connector are in scope?

2. Compare mechanisms · 8 minutes

   Where does each proposal interrupt the error path?

3. Test the washer reasoning · 8 minutes

   What does a warning allow to happen next?

4. Plan maintenance · 5 minutes

   What change would require revalidation?

Debrief

• Ask learners to identify a different failure mode outside each claim.
• Do not reward a stronger prevention label without challenge evidence.

Draw the error path, mark where the control acts and complete expected outcomes before reading the answer.

Transfer into the work

Owner: Process/design control owner with quality and operators

Record: Validated challenge matrix, response standard and change-trigger review

Review: At initial approval, defined challenge intervals and relevant changes

Evidence: Correct response to target error and credible faults without unsafe test practices

Retain containment and escalate failed controls; revise and revalidate the claim before relying on it.