Turn fishbone branches into testable hypotheses

Turn fishbone branches into testable hypotheses

The branches of a fishbone are hypotheses. Begin with one observable effect and a useful scope, then invite several explanations without promoting consensus into proof. Categories help organize discussion; they do not require every branch to be filled or make the resulting list exhaustive. Select important hypotheses and state an observation or test that could distinguish each from an alternative. Mark evidence as observed fact, hypothesis, test result or supported cause. When a test contradicts a branch, update the picture rather than defending the original workshop. A causal conclusion should explain the evidence and survive a relevant verification.

Follow the method

1. Method
2. Material
3. Measurement
4. Seam-width variation
5. All branches are unverified hypotheses

Turn three proposed causes into a discriminating test record

Fictional case: seam-width variation at Cedar Fabrication prompts three suggestions: fixture positioning differs, observers measure differently, or material thickness changes. The core fishbone preserves these three hypotheses and zero verified causes. Facilitator Elena must turn the discussion into a useful test plan without allowing agreement to become proof.

Follow the method

1. Observer technique
2. Fixture positioning
3. Material thickness
4. Overall

Three branches, zero verified causes

Convert a list of possible causes into competing, testable explanations with explicit evidence states.

Fictional case: seam-width variation at Cedar Fabrication prompts three suggestions: fixture positioning differs, observers measure differently, or material thickness changes. The core fishbone preserves these three hypotheses and zero verified causes. Facilitator Elena must turn the discussion into a useful test plan without allowing agreement to become proof.

Role: Improvement facilitator with measurement and process specialists

Normal condition

The effect has a defined scope; observations and hypotheses use different labels; a proposed test predicts what would differ under competing explanations.

The gap

The team votes fixture positioning as the root cause because it is the most familiar explanation. No comparison or corroborating record has been supplied.

• The scenario contains no authorized repair or production-setting instruction.
• A supported cause remains bounded by the conditions and uncertainty of the evidence.

1. Define the effect without a cause inside it

   Elena writes the seam-width condition, affected product/period and measurement method. She avoids “variation caused by poor setup”, which embeds the desired conclusion in the problem statement.

   Why: A neutral effect allows rival explanations to be evaluated. An imprecise effect makes nearly any story appear to fit.

   Evidence: A bounded effect statement separates what was observed from why it may have happened.

2. Keep branches provisional

   The group records the three suggestions as hypotheses and notes who can supply relevant evidence. Category labels such as method, measurement and material organize thinking but do not add proof.

   Why: The fishbone is a hypothesis organizer. The number of people agreeing with a branch is not evidence of the causal mechanism.

   Evidence: Every branch is labelled unverified; no “confirmed” color or check mark appears.

3. Choose a discriminating comparison

   For the measurement hypothesis, Elena asks qualified observers to measure the same retained seam samples under a defined method with identity concealed as appropriate. In parallel she preserves fixture and material records.

   Why: If observers disagree on the same physical sample, measurement variation deserves attention. If agreement is adequate for the task, that result weakens this particular explanation but does not prove the fixture branch.

   Evidence: The plan states rival predictions, sample identity, method and authorized owner.

4. Record evidence that could contradict the favorite

   Before results exist, the team writes that a fixture explanation would be weakened if comparable seam variation persists under verified identical positioning and measurement conditions. It identifies uncontrolled factors that could still confound the comparison.

   Why: Predictions written in advance reduce the temptation to reinterpret every outcome as support. A failed prediction is useful learning, not a reason to delete a row.

   Evidence: The test card contains both supporting and contradicting outcomes.

5. Update the state and choose the next test

   Elena leaves outcomes blank until observations are made. After a reviewed test, the group can label the evidence as supporting, contradicting or inconclusive and decide the next action.

   Why: A planned test is not a completed test. Correction or trial action must follow the evidence and relevant authority rather than the appeal of a completed-looking fishbone.

   Evidence: The record includes result, limitations, decision and next owner without invented observations.

A test result is inconclusive

Observers disagree, but the comparison used different sample locations and unclear measurement instructions.

Record the result as inconclusive for the intended hypothesis and repair the test method before attributing the effect to operators.

The comparison changed more than the intended factor. An ambiguous test cannot fairly support a specific blame or corrective action.

The next plan identifies the same measurand, location and procedure.

Run a practice session

Materials

• Blank hypothesis/test cards
• Three original branch cards
• Retained-case fact sheet

1. Write the neutral effect · 5 minutes

   Does the statement already blame something?

2. Design a rival test · 8 minutes

   What result would weaken the favorite?

3. Apply occurrence and escape · 10 minutes

   Can the carton become wrong and then escape for different reasons?

4. Debrief uncertainty · 5 minutes

   When is inconclusive the correct answer?

Debrief

• Reward an explicit evidence gap rather than a confident unsupported cause.
• Require an authorized test boundary for equipment or product challenges.

Write two rival predictions and leave results blank until evidence is supplied.

Transfer into the work

Owner: Investigation owner with process and measurement specialists

Record: Hypothesis, prediction, test result and decision log

Review: After each completed comparison

Evidence: Traceable observations and a bounded inference, including contrary evidence

Revise the test or investigate another branch; do not rename an inconclusive result as proof.