Team Operations
Root Cause Analysis: Investigate Before You Choose a Fix
Use root cause analysis to test competing explanations before choosing a fix, with an original handoff log, evidence tree and corrective-action worksheet.

What you will learn
- Write an observed problem without assuming its cause.
- Compare causal hypotheses using evidence and disconfirming checks.
- Test a corrective action and review recurrence with clear limits.
Root cause analysis investigates why a problem happened so a team can address its cause, not repeatedly treat the symptom. Start with the observed problem, gather evidence and test competing explanations. The result should be a testable explanation and a clear corrective action.
For example, a missed customer handoff might involve a reminder, an unaccepted owner, an unclear deadline or several conditions together. Sending more reminders is only a sensible fix if reminders address the cause that the evidence supports.
What are the 5 steps of root cause analysis?
Use this practical sequence: define the problem, gather evidence, identify possible causes, test the explanations, and act with a verification plan. Different RCA methods group the stages differently; these five steps provide a working structure for the exercise.
ASQ describes root cause analysis as a family of approaches and tools within a wider improvement process. Investigating a cause alone does not improve the work; a useful response must be implemented and checked.
- Define: what happened, where, when and with what consequence?
- Gather: which records and observations are trustworthy?
- Compare: what mechanisms could produce the same symptom?
- Test: what evidence supports or contradicts each explanation?
- Verify: what corrective action will be tried, and how will recurrence be reviewed?
Urgent containment can happen before the investigation is complete. If an important handoff is missing, resolve the present customer need through the authorized owner. Keep that immediate response separate from the later claim that you understand why the problem occurred.
What are examples of root cause analysis?
Consider an original fictional log of 12 customer handoffs in one week. Nine met their accepted deadline and three were late. The team initially suspects that reminders failed. The observations below are invented for learning, not Finta operational data.
| Late case | Reminder evidence | Owner and deadline evidence |
|---|---|---|
| A | Reminder delivered before the due time | No person had accepted the next action |
| B | Reminder delivered before the due time | Two teams interpreted the due time differently |
| C | Reminder delivery not established | Named owner and deadline were recorded |
The late share is 3 ÷12 =25%. That percentage describes this small log; it does not identify the cause or establish a stable failure rate. Two delivered reminders weaken the claim that every late handoff was caused by missing reminders, but they do not prove that reminders were useful or read.
The team should also inspect on-time cases. If many on-time handoffs lack an accepted owner, ownership alone may not explain the difference. A comparison that looks only at failures can miss conditions that occur throughout the process.
Build a competing-hypothesis evidence tree
| Hypothesis | Evidence that would help distinguish it | Current interpretation |
|---|---|---|
| Reminders do not reach the responsible person | Delivery records, recipient, timing and access, compared across late and on-time cases | Unresolved for C; not sufficient to explain A and B |
| The next owner is not accepted | Accepted handoff record and comparison with other cases | A candidate mechanism for A, not yet a general conclusion |
| The deadline has inconsistent meanings | Recorded timezone, cutoff and the parties’ understanding | A candidate mechanism for B requiring clarification |
| Workload exceeds available capacity | Queue, workload and accepted commitments at the relevant times | Not established by the three-case log |
Write what would make you change your mind. If every observation can be made to fit the same explanation, the hypothesis is not doing useful investigative work. Separate evidence you have from evidence you intend to collect.
You do not need to force every incident into one root cause. Multiple conditions can interact, and the most useful intervention may address a system weakness rather than identify a person to blame.
Use Five Whys without inventing answers
ASQ’s Five Whys guidance describes repeated questioning to investigate beneath symptoms and notes that it may take fewer or more than five questions. A fishbone diagram can help organize possible causes, but its branches remain candidates until checked.
For case A, ask why the action was missed, then verify whether anyone accepted ownership. If the answer is unknown, write “unknown” and identify the source to inspect. Do not continue a chain of confident assumptions until it ends at “poor communication.”
Keep the questions mechanism-focused: what condition allowed this outcome, what should have prevented it and which observation would distinguish another explanation? ASQ’s problem-solving guidance explicitly asks how a cause was verified and which other causes were studied and eliminated.
Test a corrective action and verify recurrence
ROOT CAUSE INVESTIGATION BRIEF
Observed problem, scope and impact:
Immediate containment, if needed:
Source records and coverage:
Competing explanation:
Evidence expected if true:
Supporting and contradicting observations:
Unknown and next verification:
Proposed corrective test:
Owner and authorization needed:
Recurrence measure, guardrail and review window:
What would cause us to revise the conclusion:For the fictional handoff problem, a bounded test might require explicit owner acceptance and one shared deadline definition in a small part of the process. Track whether the handoff becomes usable, whether late cases recur and whether the new check creates excessive delay. A short period with no recurrence is encouraging, not proof the issue can never return.
Use process mapping to understand the actual handoffs, and a weekly business review to keep accepted corrective actions visible. Neither replaces the causal investigation.
Copyable AI prompt
Try this with Aurora AI in Finta, Codex, or Claude to prepare your first draft.
Review [authorized problem description], [case log], [comparison cases] and [source evidence].
Build a root-cause investigation brief with competing hypotheses, expected evidence, support, contradictions and unknowns.
Recalculate rates with explicit denominators.
Propose a bounded corrective test and recurrence/guardrail checks for human approval.
Do not invent causes, blame individuals, declare an incident resolved or change source records.If the problem appears in a changing metric, explore Finta’s Metric Change Investigator for a comparison built around consistent definitions and testable explanations. Keep the case evidence beside the brief and choose the next verification from what it shows.