8 min video with captions · one-page job aid · short quiz
Job aid
From a problem to a cause you can prove
Root cause analysis finds the process that makes a problem happen, and checks it with data before anyone fixes anything. Fixing the symptom only buys time.
Step
Ask
Tool
1 State it
What is wrong, where, since when, and how much against normal?
Problem statement: no cause, no blame, no fix in it
2 Widen
What could cause it?
Fishbone: 4+ categories, 12+ causes, then choose a bone
3 Deepen
Why? And why again?
5 Whys down one branch; each answer a fact you can check; stop at a process you can change
4 Sort
What do we see, what makes it worse, what makes it happen?
Symptom · contributing cause · root cause
5 Verify
Does removing it switch the problem off?
Check the data, compare before/after rates, or run a small trial
5 Whys rules: one branch at a time. If an answer names a person ("someone made an error"), ask why the system let it happen, and keep going.
Watch out for: a cause hidden in the problem statement · stopping at "human error" · a fishbone with one busy bone · fixing before verifying.
Examples
Industry
Problem
Root cause
Verified
✈ Airline MRO
Outflow valve on ZZ-RMB replaced 3 times in 2 months (norm < 1 a year)
Task card had no torque value; not updated after a service bulletin
No fault in 2 of 3 removed valves; 90 days on the revised card: 0 removals
⚙ Manufacturing
4.2% oversize bores on CNC lathe 3 (baseline 0.8%)
No tool-life review when the bar-stock material changes
200-part trial, inserts changed at 300: 0.5%
🏦 Banking/Services
Outbound payment failures 0.6% → 1.4% of daily volume
No step to review partner change notices against our interfaces
9 in 10 rejections carry "field too long"; one-week test back to 0.6%
✚ Healthcare
12 late 08:00 doses in 2 weeks (2 before)
Delivery route changed without checking ward schedules: a support process, not nursing
n/a (supplemental example)
📦 Logistics
Mis-ships 0.3% → 0.9%, mostly two look-alike items
Slotting rules don't check for look-alike items
n/a (supplemental example)
All examples are fictional.
Quiz
Answer each question first; the answer and why open after you choose or press “Show the answer”. Nothing is saved or sent.
Pass mark: 4 of 5 questions. Q5 counts as correct at 4 of 5 rubric points.
Q1 · Airline MRO · Symptom, contributing, root
Scenario: On fictional aircraft ZZ-RMB, the cabin pressure outflow valve was replaced three times in two months for the same fault message. The team found that the connector had backed out because it was never torqued to a value. The task card says only "secure connector": it was written before a service bulletin added a torque value. The repeat-defect history is not shown to the next shift.
Question: Which item is the root cause?
Answer: C
Why:
A) This is the symptom: what everyone sees.
B) This is a contributing cause: it made each replacement look like a first occurrence, but removing it alone wouldn't stop the connector backing out.
C) ✓ Fix the card, and the connector is torqued every time: the problem stops coming back. It's a process you can change.
D) "Technician error" is never the end of the ladder. Ask why the system let it happen: the card gave no torque value.
Q2 · Manufacturing · Verify the cause
Scenario: Since week 10, 4.2% of shafts from fictional CNC lathe 3 have an oversize bore, against a 0.8% baseline. A harder bar stock arrived in week 10. Inserts (the cutting tips) are still changed at a fixed count of 500, a count set for the old material. The team thinks the root cause is that tool life isn't reviewed when the material changes.
Question: Which check would best verify that root cause?
Answer: B
Why:
A) This tests the gauge, a different cause on the fishbone, not tool life.
B) ✓ It switches the suspected cause off in a small trial. If the rate drops toward the baseline, the cause is confirmed. (In the example, 1 of 200 was oversize: 0.5%.)
C) It's a fix aimed at a person, and it tests nothing about tool life.
D) It's a fix, not a check, and it avoids the real gap: tool life isn't reviewed when material changes.
Q3 · Banking/Services · Problem statement
Scenario: A fictional bank's payments team is writing up a rise in failed outbound payments.
Question: Which problem statement is complete and measurable?
Answer: B
Why:
A) It names a cause and blames a team: that closes the search before it starts.
B) ✓ It has all four parts: what (failures), where (mainly one partner network), when (since 3 September), how much (1.4% against 0.6%).
C) It has no date or figures, and it already contains a solution.
D) It has no baseline, no date and no location.
Q4 · Logistics · Transfer question
Scenario: At a fictional warehouse, wrong items sent to customers ("mis-ships") rose from 0.3% to 0.9% of orders in October, mostly for two products that look almost the same. Pickers took the neighboring product because the two were stored side by side. That happened because the storage plan ("slotting": deciding where each product goes) was redone by product size, and the slotting rules don't check for look-alike items. October was also the busiest month of the year.
Question: Which is the root cause?
Answer: D
Why:
A) This is the symptom.
B) A contributing condition: busy months make mistakes more likely, but don't create the look-alike problem.
C) This is a step on the ladder, not the end, and it points at people. Ask why the system made it easy.
D) ✓ Change the rule, and look-alike items are never stored side by side again.
Why this transfers: logistics isn't planned for M5, and the scenario explains its terms (mis-ship, slotting). A learner who can sort symptom, contributing cause and root cause can answer without warehouse knowledge.
Q5 · Build exercise (learner's choice)
Task: Pick a problem that keeps coming back in your own work (or use this case: a team's weekly report goes out late 3 weeks in 5 since a new data system went live in May; before that it was late about once a quarter). Write:
A problem statement in four parts.
A fishbone with at least 4 categories and at least 12 possible causes.
One 5 Whys ladder down one bone.
How you would verify the root cause with data before fixing it.
Rubric (1 point each; pass at 4 of 5):
#
Point awarded when…
1
The statement has what, where, when and how much against a baseline, with no cause, blame or solution in it
2
The fishbone has at least 4 categories
3
The fishbone has at least 12 distinct, plausible causes
4
The ladder ends in a process that can be changed, not a person
5
The verification uses data or a trial that would switch the problem off and on
All examples are fictional.
Cookies on InLine Ops. We'd like to use optional analytics cookies to see how the site is used and improve it. They're only set if you accept. Read more about cookies
ACADEMY · IN PRODUCTION
The Academy opens soon
We use your email only to tell you about this. See our Privacy Policy.