How Dr. Thomas Krause Thinks About 4 Conditions Before an Incident Investigation Begins

An incident investigation can become a blame exercise before the first interview starts. The team arrives with a settled story, the operator is treated as the main explanation, and the review then collects statements that confirm what leaders already believe.

Dr. Thomas Krause challenged that pattern on Episode 11 of the Headline Podcast. He explained that incident analysis often appears to prove employee fault, even though the difficulty of following the procedure may have been created by decisions made months or years earlier. That observation changes the first task of an investigation. Before asking who made the final error, leaders need to test whether the work was set up to succeed.

A decision-ready incident investigation begins before the interview, because the quality of the evidence depends on the work conditions, the operating decision, the procedure, and the leadership signals that shaped the event.

Why the interview should not be the starting point

Interviews matter, but they are not a substitute for reconstructing the work. A worker can describe what happened from one position in the system, while the investigation must explain how the task, equipment, schedule, supervision, and controls interacted.

When the interview comes first, the question wording can narrow the explanation. “Why did you not follow the procedure?” assumes that the procedure was usable, available, understood, and compatible with the conditions on the day. The better opening asks what the job required, what changed, and which decision became difficult.

James Reason’s work on latent failures gives investigators a useful discipline. A visible action may be close to the event, while the conditions that made that action likely were created elsewhere. That does not remove personal accountability. It makes accountability more accurate.

Condition 1: define the decision that failed

Every investigation should name the safety decision that did not hold. The decision might have been to isolate energy, delay a lift, stop a line, verify a barrier, escalate a deviation, or refuse a restart.

Without that definition, the review becomes a story about behavior. With it, the team can ask which information was available, who owned the decision, what authority existed, and how much time the operation allowed.

Imagine a maintenance crew that resumes work after a late design change. If the investigation only records that a technician missed a step, it leaves the central decision invisible. The more useful question is why the change was accepted without a new control review and why the person closest to the hazard had no clear route to pause the job.

That framing also creates a practical test for leaders. If the same decision would be difficult for a competent person on the next shift, the remedy must change the work rather than only remind people to be careful.

Condition 2: preserve the work as it was performed

Procedures describe intended work. An investigation needs evidence of performed work, including the sequence, interruptions, handoffs, tools, alarms, staffing, and conditions that existed at the time.

Start with time-stamped records, permits, control-room logs, photographs, equipment states, maintenance history, and the actual version of the procedure. Then compare those records with the task as people experienced it. A document that was technically correct but unavailable at the point of use is not an effective control.

The comparison should remain factual. It should not label the gap as noncompliance before the team understands why the gap existed. Production pressure, a conflicting instruction, a missing part, a poor interface, or an unclear handover can each change the decision space without appearing in the final incident form.

The Evidence Chain guide on Headline Podcast’s blog provides a useful companion lens, because an investigation loses credibility when observations, records, interpretations, and decisions are mixed together. Keeping those layers separate makes the final explanation easier to challenge and improve.

Condition 3: test whether the procedure could work

A procedure can be correct on paper and impractical in the field. That distinction matters because an investigator who treats the document as proof of control may mistake a design failure for a worker failure.

Ask whether the procedure matched the equipment, the sequence, the environment, and the competence available during the task. Check whether it required simultaneous actions that one person could not safely perform. Review whether the language identified the critical point clearly enough for a worker under time pressure.

Dr. Krause’s podcast observation is especially relevant here. If following the procedure was made difficult by earlier organizational decisions, then repeating the procedure in training will not close the gap. The investigation must identify the decision that made faithful execution unrealistic.

This is where leaders should resist the comfort of a quick corrective action. A refresher briefing may be necessary, but it is weak when the underlying sequence, staffing, equipment, or approval path remains unchanged.

Condition 4: examine the leadership signal around bad news

People report what they believe leaders can hear. When a site rewards speed, treats deviations as incompetence, or praises clean statistics without asking how the work is controlled, the investigation may receive a polished version of events.

That signal can appear before the incident. A supervisor may discourage escalation because the job is already late. A manager may ask for a “simple explanation” because a complex one threatens the schedule. A safety professional may soften a finding because previous reviews punished the messenger rather than the condition.

Rodney Rocha’s Headline Podcast discussion on bad news reaching leaders offers a related test. The first missed signal is often not a formal refusal. It is the silence that appears when people expect fear, ridicule, or retaliation after they raise a concern.

Investigators should therefore ask who knew about the precursor, what happened after it was raised, and whether the response encouraged the next report. The answer reveals whether the organization is learning from weak signals or teaching people to keep them private.

What leaders should ask before approving corrective actions

Corrective actions should address the decision conditions that the investigation has demonstrated. A strong review does not need a long list of actions. It needs actions that change exposure, authority, information, or control reliability.

• Which decision must become easier or safer on the next shift?
• Which condition made the original decision difficult?
• Who owns the revised control and how will that person verify it?
• What evidence will show that the work changed rather than the paperwork?
• What happens when the revised control is unavailable or ineffective?

Use the answers to reject weak actions that only restate the expected behavior. “Be more careful” is not a control. “Retrain everyone” is not a complete response when the procedure, interface, staffing, or authorization path created the difficulty.

Andreza Araujo’s safety-culture work consistently places attention on the gap between declared commitment and operated conditions. That gap is where an investigation either creates credibility or loses it. Leaders show the value of the review through the changes they are willing to make after the report is written.

How to keep accountability without reducing the explanation

System conditions and individual choices are not competing explanations. A person can make a decision that requires accountability, while the organization can still be responsible for the conditions that shaped the decision.

Use a layered conclusion that distinguishes the action, the immediate conditions, the contributing conditions, and the governance decisions that allowed the exposure to remain. That structure avoids two common errors. It does not excuse every action, and it does not pretend that discipline alone will prevent recurrence.

The conclusion should also identify what a competent worker would have needed to make the safer decision. If the answer includes information, time, authority, equipment, or supervision that was absent, the corrective plan must supply it.

Recommendation

Before the first interview, define the failed decision, preserve the work as performed, test whether the procedure was usable, and examine the leadership signal around bad news. These four conditions determine whether the investigation will explain the event or simply assign it.

Dr. Thomas Krause’s message is not an argument against accountability. It is an argument for placing accountability where decisions are made. When leaders investigate the conditions that shaped the work, they can protect people, improve controls, and still identify choices that require correction.

Use the next incident review to compare the written procedure with the task as performed, then ask which organizational decision made the gap possible. That question is more demanding than asking who missed a step, but it is also more likely to produce a control that survives the next shift.

For a practical follow-up, review the evidence chain for a decision-ready investigation and compare it with the nine-step near-miss review process. Leaders who want to strengthen the reporting climate can also use the bad-news-to-leaders discussion as a companion reading.