How to Run a Safety Culture Evidence Walk in 8 Steps

A safety culture survey can tell leaders that people feel positive, but the field can still show missed handovers, silent workarounds, and supervisors who only appear after the job is already drifting. A safety culture evidence walk exists to close that gap, because culture is not the answer people give in a meeting room. Culture is what leaders can verify where the work happens.

A safety culture evidence walk is a short, structured field review in which a leader checks visible traces of planning, supervision, speaking up, stop-work behavior, and follow-through, then compares those traces with what the organization claims in its reports.

Across 25+ years of executive EHS work and more than 250 cultural transformation projects, Andreza Araujo has seen that a site learns more from a disciplined walk than from a polished presentation when the walk is built around evidence. As Andreza Araujo writes in A Ilusão da Conformidade ("The Illusion of Compliance"), the real test of a system is what happens when nobody is watching, and that is exactly why this method starts in the field, not in the dashboard.

What you need before starting

Pick one area, one question, and one leader who can act on what is found. A culture walk fails when it turns into a general tour, because people start to perform for the visitor instead of exposing the operating reality. Keep the visit short enough that the crew can stay in normal work mode, and decide in advance what evidence counts as real.

The right starting question is not "Do people like working here?" The better question is, "Does the worksite show the same discipline that the reports describe?" If you want a practical reference point, compare the walk with the evidence categories in Safety Culture Traces Explained: 5 Evidence Types Leaders Can Verify, because the walk and the article should point to the same kind of proof.

James Reason remains useful here because his layered view of failure reminds leaders that one clean conversation does not prove a strong system. The walk should search for weakened layers, not for slogans.

Step 1: Define the one operating question

Start with a single question that can be answered in the field. For example, you may want to know whether the crew starts high-risk work with a real pre-task review, whether supervisors close the loop after a concern, or whether the area owner can explain the controls without reading a script. One question keeps the walk sharp, while three questions usually dilute the observation.

The leader should write the question in plain language before entering the area, because the question determines what evidence matters and what noise gets ignored. If the team cannot turn the question into a visible check, the walk is too abstract to be useful.

Step 2: Choose the route and the proof you will accept

Select the exact location, the shift, and the moment in the work cycle you want to observe. A pre-start briefing shows a different set of traces from a mid-shift recovery or a shift handover, so the route should match the decision you want to test. If the issue is control quality, go where the control is supposed to be used. If the issue is voice, go where silence would normally cost something.

The walk is the only one of the three that can show whether the story and the work still match. That is why it is a strong complement to survey results, not a replacement for them.

Step 3: Set the evidence standard before you walk

Decide which traces count as proof. In a safety culture walk, proof is usually visible and ordinary: a supervisor who asks a real question, a permit that matches the job front, a boundary that is actually marked, a stop-work decision that was respected, or a handover that includes the risk that still remains. Without a standard, the leader will notice only the obvious items and miss the operating detail that shows whether culture is real.

Do not accept polished language as evidence. As Andreza Araujo argues in A Ilusão da Conformidade ("The Illusion of Compliance"), the fact that a control looks complete on paper does not mean the worksite is being controlled. The evidence standard should force the leader to see whether the barrier is being used, not only whether it was documented.

Step 4: Walk with one supervisor and one operator

Bring the smallest group that can still answer the question. One supervisor and one operator are often enough, because the point is not to produce a crowd. The point is to compare what management believes with what the people closest to the task can actually show.

Ask the crew to demonstrate, not to describe. If the work involves a pre-task review, ask where it happens and what gets checked. If the risk involves isolation, ask what would prove the equipment is truly safe to touch. If the issue is psychosocial, ask what would make a worker feel safe enough to raise a concern before the job starts. The answer should appear in the field, not only in the vocabulary.

Step 5: Compare words to work

This is the most important point in the walk. When the answer sounds disciplined but the field looks improvised, the organization has found a culture gap. The goal is not to shame the crew. The goal is to show where the system still depends on memory, personal caution, or the luck of having an experienced person nearby.

Use the mismatch as a signal. A clean form, a good slogan, and a confident explanation are weak evidence if the barrier is missing or the supervision arrives late. The walk only becomes useful when the leader names the gap in concrete terms, because vague praise is one of the easiest ways to protect a weak process.

Step 6: Translate each gap into one owner and one fix

Every gap should leave the walk with an owner, a due date, and a next check. If the problem is a missing handover step, the owner may be the shift supervisor. If the problem is that operators cannot explain the control, the owner may be the area leader. If the problem is that the worksite keeps improvising around the same barrier, the owner may be operations management. The point is to assign responsibility where the fix can actually happen.

Do not let the walk dissolve into a lesson with no action. A culture walk that generates only general encouragement changes nothing, while a culture walk that creates one specific operating correction starts to change the next shift. That is the practical difference between observation and management.

Step 7: Close the loop within seven days

Return to the same place within a week and check whether the evidence changed. The follow-up walk is where credibility is won or lost, because people quickly learn whether the leader intended to act or only intended to be seen. If the fix was real, the field should look different. If it was not, the original gap will still be visible.

This is where James Reason helps again, because the second visit shows whether a layer was strengthened or only discussed. A weak site often has many good intentions, but intention does not block risk. Evidence does.

Step 8: Make the walk part of the monthly rhythm

One walk is useful. A repeated rhythm changes culture. Keep the format stable enough that the team knows what will be checked, then rotate the area, the shift, and the question so the organization does not learn how to game the visit. Over time, the walk becomes a management habit that exposes drift before it hardens into normal work.

Andreza Araujo has written in Cultura de Segurança: Da Teoria à Prática ("Safety Culture: From Theory to Practice") that culture is built through repeated decisions. A monthly evidence walk fits that principle because it turns leadership into a visible discipline instead of a speech. When leaders show up, ask one real question, and verify the answer in the field, the organization starts to treat evidence with more respect than appearance.

For plant managers and supervisors, the test is simple. If the evidence walk keeps finding the same gap, the site does not have a communication problem. It has an operating problem, and the next conversation should be about the barrier that still is not working.