To find the root cause of a recurring problem, describe what keeps happening, collect examples, and test the conditions that allow it to happen again. The first plausible explanation is a starting hypothesis. It becomes useful when evidence supports it and changing it reduces the problem.
Repeated reminders, apologies and emergency fixes can keep work moving. But if the same failure returns, the process deserves a closer look.
Separate the symptom from the mechanism
A puddle is a symptom. A leaking joint may be the immediate cause. A badly supported pipe, an unsuitable fitting or an inspection gap might explain why the joint repeatedly fails.
The point is not to keep digging forever. It is to find a cause you can understand well enough to act on, while recognising that several causes may interact.
The Institute for Healthcare Improvement’s Five Whys guidance warns against rushing to a solution before understanding a problem and notes that there may be multiple root causes.
Start with a problem statement you can observe
“Our team is careless” is a judgement. “Four of the last ten orders were sent without the requested attachment” is a statement you can investigate.
Include what happened, where, when and how often. Describe the expected result too. If you do not know the frequency, collect a small record before presenting an estimate as fact.
| Vague statement | Investigable version |
|---|---|
| Meetings are useless. | Three recent meetings ended without an owner or due date for the agreed task. |
| The system always breaks. | The upload failed on three files above a certain size this week. |
| Nobody communicates. | Two handovers omitted the customer’s latest requirement. |
These are invented examples. Replace them with your own observations rather than borrowing the diagnosis.
Look at successful cases as well as failures
Compare a failed order with one completed correctly. Was the same template used? Did the requirement arrive at the same stage? Did the person have the same information and time?
Differences give you possible explanations. Similarities can rule out an easy story. If the same person completes most orders correctly, “they need to care more” explains very little about the exceptions.
Keep a simple evidence note: date, situation, expected outcome, actual outcome and relevant conditions. Where appropriate, use logs or records rather than memory alone. Protect private information while investigating.
Use “why” as a question, not a script for inventing answers
The Five Whys technique repeatedly asks what led to the previous answer. ASQ explains that it can take fewer or more than five questions. The number is a prompt to explore, not proof that you have reached a final cause.
Consider a fictional missing-attachment problem:
- Why was the attachment missing? The sender used an old email template.
- Why was that template used? It was the first result in the shared folder.
- Why was an obsolete template still presented as current? Old and current templates shared the same location and similar names.
- Why did the process not catch it? There was no clear owner for retiring templates or checking required attachments.
Every answer needs support. Inspect the file history, folder and workflow. If you do not know an answer, write “unknown” and decide how to investigate it. A neat chain of guesses is still a guess.
Branch when the evidence points in different directions
Perhaps old templates explain some missing attachments, while other cases happen because the customer sends a requirement after dispatch. Do not force both into one story.
List the branches separately. For each one, record what you observed, what remains uncertain and what would distinguish it from competing explanations. In complex systems, a qualified investigation may need more than a short questioning exercise.
Useful questions include: “What evidence would make us reject this explanation?” and “When does the problem not occur?” They help prevent a favourite theory from becoming the answer by default.
Distinguish containment from prevention
Containment deals with the current consequence: send the missing document, correct the order or stop using faulty equipment. Prevention changes the conditions that produce repeat failures.
Both can be necessary. In the template example, a preventive change might be one clearly marked current template, archived obsolete versions and a required attachment check before sending. The right fix follows the confirmed mechanism.
“Be more careful” leaves the process largely unchanged. Ask what makes the correct action easier and what catches an omission before it reaches the customer.
Test a manageable change and watch for side effects
Choose one low-risk change, assign an owner and define what improvement would look like. For example, try the revised template process for the next twenty eligible orders and record missing attachments and time spent preparing each order.
A small trial provides evidence for the tested conditions, not a universal guarantee. Check whether the mix of orders changed, whether people followed the new process and whether another problem appeared.
If results improve, continue monitoring after the initial attention fades. If they do not, revisit the explanation instead of immediately blaming execution.
A reusable root-cause note
- Problem: What observable result is recurring?
- Evidence: Which examples and records support that description?
- Possible causes: What mechanisms could produce it?
- Checks: How will you test each explanation?
- Action: What change addresses the supported cause?
- Review: Who will measure results, and when?
Common questions
Is there always one root cause?
No. Several conditions can combine. Keep the explanation as complex as the evidence requires.
Must I ask why exactly five times?
No. Stop when you have a supported, actionable understanding, or continue investigating when uncertainty matters.
Can I use this for a serious safety incident?
A short guide is not a complete incident investigation. Follow the relevant safety procedures and involve qualified people. Address immediate hazards before experimenting.
Related: Our Wi-Fi troubleshooting guide demonstrates how comparisons can narrow a problem, while checking an AI answer explains why a convincing explanation still needs evidence.
