5 Whys with evidence and alternative branches

5 Whys with evidence and alternative branches

Begin with a specific observed gap. For each proposed cause, ask what evidence supports the link and what observation would contradict it. Keep alternative explanations visible when the evidence branches. The number of questions is less important than reaching a cause that can be tested and acted on. Do not stop at a label such as human error, and do not invent certainty to complete a chain. A countermeasure should change the mechanism and be checked for effectiveness. Complex interacting variables may require another analysis method. The diagram supports disciplined reasoning; it is not proof by itself.

Follow the method

1. Observed gap
2. Hypothesis A
3. Hypothesis B
4. Discriminating test
5. Decision

Test two explanations before closing the why chain

Fictional service team: three change-request packets arrived without a required drawing. Reviewer Sol is asked to finish a five-whys form that already ends with staff not paying attention. The team can inspect packet histories and reproduce submission in a training environment.

Follow the method

1. A: omission accepted at entry
2. B: attachment lost later
3. Staff inattention
4. Next decision

The missing attachment is an event, not an attitude

Build a causal explanation with evidence at each link, retain a competing explanation and define a test whose result could change the decision.

Fictional service team: three change-request packets arrived without a required drawing. Reviewer Sol is asked to finish a five-whys form that already ends with staff not paying attention. The team can inspect packet histories and reproduce submission in a training environment.

Role: Process owner and frontline colleague investigating the specific omission.

Normal condition

Every submitted change packet includes the applicable drawing or an explicit approved exception.

The gap

Three recorded omissions; no evidence yet connects them to attention.

• Use synthetic test packets; do not alter live customer requests.
• A plausible chain is a hypothesis until its links are supported; five is not a mandatory stopping count.

1. State the event precisely

   Sol records packet identities, required drawing and point of detection. He asks where the attachment should first enter the process and separates that point from review.

   Why: Investigating where the omission was found can miss where it arose.

   Evidence: Event statement and expected handoff are explicit.

2. Make each why a supported link

   The team asks why a packet could arrive without the drawing and labels each proposed answer observed, hypothesized or unknown. It does not copy staff inattention into the cause field.

   Why: A judgment about people does not describe a mechanism that can be checked.

   Evidence: Portal acceptance and transfer loss remain separate hypotheses.

3. Design a discriminating observation

   Use a training packet without a drawing to test whether the portal accepts submission; separately send a packet with a uniquely identified test drawing and trace whether it survives each handoff.

   Why: The two tests distinguish omission at entry from loss after entry. One generic successful submission cannot distinguish them.

   Evidence: Before testing, record predicted results for each explanation and retain actual results separately.

4. Follow the result instead of the form

   If the portal rejects an empty attachment, revise the entry hypothesis rather than adding more unsupported whys. If a trace loses a drawing, inspect that precise transition and test the proposed change.

   Why: Contrary evidence narrows learning. A cause claim needs a verified mechanism and an effective response, not a filled chain.

   Evidence: The completed planning record leaves results open and defines the next decision for each outcome.

Both mechanisms occur

In a fictional follow-up, one route permits an empty packet and another loses an attachment during conversion.

Keep two evidenced branches and their affected scopes; test responses for both. Do not force all events into the first explanation.

A recurring outcome may have multiple mechanisms.

Link each reproduced event to route, condition and observation; verify occurrence and escape controls separately.

Run a practice session

Materials

• Event and hypothesis cards
• Blank cause-test record
• Synthetic packet example

1. Separate facts and labels · 5 minutes

   Which claim could another observer verify?

2. Build competing test plans · 7 minutes

   What result differs between A and B?

3. Practice acknowledgment case · 8 minutes

   What would disprove your explanation?

4. Debrief branches · 5 minutes

   Must there be one cause?

Debrief

• Ask for a mechanism when a learner names attitude.
• Do not reward chain length over tested reasoning.

Write the competing prediction before imagining any result, then check the suggested reasoning.

Transfer into the work

Owner: Process owner with people performing the affected handoffs

Record: Evidence-labelled cause/test record

Review: After each test and after the proposed response is trialed

Evidence: Reproducible mechanism, contrary evidence and recurrence/escape verification

Retain uncertainty, improve observation or use a more suitable analysis method.