Business

Root cause vs contributing factor: compare the evidence, not the labels

Stefan Weiss, Co-Founder & CEO

Stefan Weiss

Root cause vs contributing factor: compare the evidence, not the labels

A root cause is an underlying condition that an investigation identifies as a source of a problem. A contributing factor helps the problem occur, makes it more likely or increases its impact. A trigger is the event that sets a particular occurrence in motion.

Those descriptions are useful starting points, but the boundary is not universal. Different investigation methods use the terms differently, and a complex failure can have several interacting causes. The practical test is whether you can explain how a condition affected the outcome, show the evidence and choose a change that addresses that mechanism.

Do not downgrade a condition merely because somebody calls it “contributing”. It may still deserve an important corrective action. Equally, calling an explanation “the root cause” does not make it verified.

This guide compares the terms through an ordinary project-delivery example. It is not a substitute for a trained safety, legal, clinical or regulatory investigation.

The difference in plain language

The American Society for Quality defines a root cause in terms of a factor behind a nonconformance that should be eliminated through process improvement. Its guidance also describes root cause analysis as a family of approaches, rather than a single universal technique.

The Canadian Centre for Occupational Health and Safety distinguishes direct and indirect causes and notes that indirect causes may be described as contributing factors. It emphasises looking beyond immediate findings. These definitions help explain why two competent reviewers may label the same condition differently.

For a practical review, agree what the terms mean in your team before using them to rank explanations. Then attach each explanation to an observed event or condition. Avoid starting with an empty “root cause” box and pressuring people to fill it with one short phrase.

Trigger: what happened at this point?

A trigger helps locate the event in time. Someone selected an old attachment. A service restarted. A customer changed a requirement. These events matter to the timeline, but they may leave a deeper question unanswered: why could this action create this result?

A trigger can be part of a causal explanation without being its only useful level. Removing a particular triggering event might prevent this occurrence while leaving the process exposed to a similar one.

Contributing factor: what made the result more likely or worse?

A contributing factor might be a confusing file name, a late handoff, poor visibility of an outstanding approval or a recovery process that depends on one person. State its mechanism: a confusing name made the obsolete version difficult to distinguish from the current one.

“Communication was poor” is too broad to evaluate. “The revised approval was sent in a separate thread that the sender of the customer email had not received” gives the investigation something specific to check.

Root cause: what underlying condition needs to change?

An underlying condition may sit in the workflow, system design, responsibilities or management of change. For example, the release process may allow an external document to be sent without a designated current version or a required approval check.

That is still a hypothesis until evidence supports it. A detailed process explanation can be wrong. Check how the process actually operated, including exceptions, before presenting it as the investigation’s conclusion.

A worked example: the wrong project plan reaches a customer

Imagine a fictional project team that sends a customer an obsolete delivery plan. The customer begins arranging work around the wrong dates. The team discovers the error the next morning and issues a correction.

The initial explanation is: “The account manager attached the wrong file.” This describes an action close to the event. It does not yet explain the circumstances that made the error possible or why the check before sending did not catch it.

Suppose the review establishes the following sequence. A revised plan was approved in one message thread. The shared folder still contained two files called “final”. The person sending the email used a link from an earlier handoff. The release instructions did not identify who had to verify the externally shared version. A colleague noticed the mismatch only after the customer replied.

The example contains several candidate explanations, each with a different role. Selecting the old attachment is the immediate action. Ambiguous file names and a stale handoff link are contributing conditions. A process with no controlled current version or named external-release check is a possible underlying weakness. The delayed detection increases the duration of the customer’s exposure to the wrong plan.

These roles describe an argument to investigate. They do not prove that all five conditions actually occurred in a real incident. A review needs the messages, file history, permissions and relevant accounts before adopting that account of events.

Compare explanations using three checks

1. State the mechanism

Ask: how would this condition produce or worsen this outcome?

For the ambiguous file names, the mechanism is that the sender could not reliably distinguish the approved version from the obsolete one. For the missing release check, the mechanism is that no required verification prevented the obsolete plan from leaving the team.

Compare that with “the team was busy”. Workload may matter, but the explanation needs a connection. Did the handoff arrive after the usual review window? Was the reviewer unavailable? Was a specific control skipped? Do not turn a plausible background condition into an established cause without that link.

2. Identify evidence that could support or weaken it

A statement in an interview is evidence of what that person recalls or reports. A message timestamp may establish when a message was sent. Neither automatically proves what another person saw, understood or acted on.

For the stale-link explanation, inspect the link actually used and the version it resolved to at the relevant time, where that history is available. For the approval explanation, compare the approved version with the attachment. For the missing-check explanation, read the documented procedure and ask how the team normally applies it.

Also look for evidence against the explanation. If the sender opened the current file and then deliberately chose a different attachment, ambiguous naming may explain less than first assumed. If a release check existed but was not usable under the circumstances, “no procedure” is an inaccurate conclusion.

3. Connect the explanation to a proposed change

Ask what would change if you addressed this condition, then be specific about the remaining risk. A single designated external version could reduce version confusion. An approval record tied to that version could make verification clearer. Neither is proof that every future handoff will succeed.

A change can address a contributing factor and still be worthwhile. Renaming files may be a useful immediate step even while the team redesigns the release process. Avoid making the investigation a contest to find the one label that justifies all the action.

Why “human error” is an incomplete stopping point

It can be accurate to describe a mistaken action. The problem is treating that description as the end of the explanation.

OSHA’s incident-investigation guidance advises looking beyond immediate causes and warns against stopping at carelessness or failure to follow a procedure. In an office review, the corresponding questions are practical: what information was available, what options were visible, what checks existed and why did they fail to prevent or detect the problem?

This does not erase individual actions or accountability. Record what happened accurately and follow the appropriate process for misconduct or other serious concerns. The learning review still needs to understand the conditions around the action rather than replacing them with a broad character judgement.

In the fictional plan example, “be more careful” leaves the two indistinguishable final files and the unclear release responsibility in place. A more useful proposal changes those conditions and defines how the team will check whether the change works.

Use a timeline before choosing a method

Start with the outcome and work backwards through the relevant events. Separate established facts, reported accounts, interpretations and unknowns. Keep the source beside each important claim so another reviewer can inspect it.

If two accounts conflict, preserve both until further evidence resolves the difference. One person remembering approval on Tuesday and another remembering it on Wednesday is a discrepancy to investigate, not a licence to choose whichever fits the emerging story.

The Humanitarian Innovation Guide’s diagnosis guidance likewise places analysis and interpretation after information collection. That sequence matters: a diagnosis should answer to the evidence rather than organising the evidence around a preferred answer.

Where five whys and fishbone analysis help

A five-whys discussion can help a team move past the first explanation by repeatedly asking what allowed a condition to exist. It becomes weak when each answer is an unsupported guess or when the facilitator insists on exactly five levels. Treat each answer as a claim to test, and branch the discussion when several conditions interact.

A fishbone analysis can help collect candidate factors across different aspects of the system. Its breadth is useful for discovery, but placement on a diagram does not establish causation. Check the strongest candidates against the timeline and the evidence afterwards.

CCOHS lists several investigation methods and states that there is no single definitive way to conduct root cause analysis. Use the method required by your organisation where one applies. For an ordinary project review, select a method that helps expose and test explanations rather than one that merely produces a tidy picture.

Compare recorded accounts without asking AI to prove a cause

Saved conversations can help reconstruct what people described, which decisions they discussed and which questions remain unanswered. In Weeve’s chat with recordings, select the relevant project and recordings before requesting a comparison.

A useful request is: “Compare these accounts of the plan handoff. Separate exact reported events from interpretations, keep conflicting accounts visible and cite the supporting passages. List questions that need evidence outside the recordings.”

Inspect the citations and the amount of material reviewed. A missing mention in a few retrieved passages does not establish that an approval never happened. A summary should not silently turn a suggestion into an agreed control, or a person’s recollection into verified file history.

The scope selector illustrated here is genuine approved synthetic UI. It demonstrates narrowing the material for a comparison. It does not show an incident investigation, a verified causal model or a completed analysis.


Genuine Weeve recording selector with a synthetic Product planning conversation selected

Choose corrective actions at the right level

For each supported explanation, describe the proposed change, its owner, the condition it addresses and how you will review the result. Keep immediate containment separate from prevention. Correcting the customer’s plan contains this incident; controlling the approved external version addresses part of the future process.

Avoid promising elimination of a problem from one action. Check whether a new control is usable, whether people know when it applies and whether it creates a new failure point. If the designated approver is absent, for example, the process needs a clear alternative rather than an informal bypass.

Review the first relevant uses after the change. Did people identify the right version? Was approval visible before external release? Did exceptions follow the intended route? These checks are more informative than recording that a new procedure was written. A quiet period with no reported recurrence is useful context, but it may not establish effectiveness when the underlying event is rare or detection is weak.

A conclusion another reviewer can challenge

A strong review conclusion identifies the outcome, the important conditions, the evidence supporting each connection and the limits of the analysis. It also states what the team proposes to change and what remains uncertain.

For the fictional example, a careful conclusion might be: “The old plan was attached using an earlier handoff link. Two final-labelled versions were available, and the reviewed release instructions did not identify a mandatory version check. We propose one designated external version and a recorded approval before release. We still need to verify how the sending tool handled the attachment.”

That is more useful than “root cause: human error”. It gives the team an explanation to test and a change to review. The labels support that work; the evidence and the resulting decisions carry it.