When Culture Scores Rise but Field Decisions Do Not: Lessons From 250+ Projects

A site celebrates after its safety culture survey improves by 12 points. Leaders share the result in the town hall, the dashboard turns green, and the transformation is marked as a success. Three weeks later, a supervisor allows a permit exception because the crew is already behind schedule.

That sequence does not prove the survey was wrong. It proves that a score and a decision measure different layers of organizational life. The score can describe what people believe is changing, while the permit decision reveals what the operation still rewards under pressure.

A safety culture improvement is credible when better perceptions are followed by better decisions in the work. Across more than 250 cultural transformation projects, Andreza Araujo's practical test has been to connect what people report with what leaders authorize, stop, escalate, resource, and verify after the meeting ends.

Initial scenario: the score improved before the work did

Leaders can invest seriously in listening, interviews, commitments, and employee feedback, yet still treat a favorable response as the final product. A team may become better at describing the desired culture without gaining the authority, time, or support required to operate it.

This is the distinction Andreza Araujo develops in Safety Culture: From Theory to Practice. Culture is not only the language people use about safety. It is the pattern of decisions that survives contact with cost, delay, hierarchy, and uncertainty.

The decision that exposed the gap

The gap became visible when a routine control was challenged during production pressure. The supervisor knew the expected standard, the crew knew the concern, and leadership had recently reported a stronger safety climate. Still, the job continued with a temporary exception that had not been properly reviewed.

The important question was which decision path made the exception easier than the escalation. James Reason's work on latent conditions helps explain why this matters, because the visible deviation is often the final expression of a weakness in authorization, workload, resources, or supervision.

The intervention: move from sentiment to decision evidence

Leaders changed the review question. Instead of asking whether people felt more committed, they asked which safety decisions had changed since the previous review and what evidence supported the claim.

Each team mapped when a worker could stop the job, when a supervisor had to escalate, who could accept a temporary change, how an action owner received support, and how the team would verify the control after closure. A survey concern that could not reach a decision owner was treated as incomplete.

The approach reinforced the principle in Make the Difference: Be a Leader in Health & Safety. Leadership is visible when the organization responds consistently to the same risk, even when the response costs time or interrupts the plan.

Execution: change the weekly conversation

The weekly review moved away from campaign updates and toward a short evidence cycle. Leaders selected one concern, traced it to the decision owner, checked the field condition, and asked the supervisor what made the response easy or difficult.

A good score with no changed decision required investigation. A lower score alongside faster escalation could show that the workforce had become more willing to describe reality. Leaders stopped treating discomfort as regression when it exposed a condition that had previously stayed hidden.

The review separated two questions. Can people speak? Can the organization act? A mature culture needs both, because a report without response teaches silence and a response without voice leaves leaders blind.

Measured result: the evidence became more honest

The result was not a single headline reduction. The transformation produced a more reliable view of where risk was managed and where it was absorbed. Leaders saw more explicit escalations, clearer ownership, fewer vague closure statements, and stronger links between employee concerns and field verification.

Across the 250+ project portfolio, Andreza Araujo's experience supports a practical conclusion. The strongest evidence of cultural change is not that employees describe leadership more positively. It is that leaders make different decisions when the work becomes inconvenient.

Generalizable lessons from the case

Do not ask a survey to prove what only field evidence can prove. Perception data can identify trust, confidence, or concern, but it cannot demonstrate that a critical control is operating.

Preserve the connection between a concern and a decision owner. Review exceptions because they reveal the real boundary, and treat follow-up as part of the culture rather than as administration after the culture work.

What leaders should apply in the next 30 days

Put one recent culture result beside one operational decision that occurred during pressure. If the score says trust is rising but the decision shows that concerns are still dismissed, use the contradiction to choose the next intervention.

Select three decision moments for weekly verification, such as a stop-work call, a permit exception, and a corrective-action closure. Identify who can act, the support required, the response time, and the evidence that will confirm the change.

Use the safety culture ownership review as a companion diagnostic, and compare it with the weekly leadership decision cadence.

Conclusion: culture is proven at the point of decision

A safety culture score is a useful signal, but it is not proof of safer work. The proof appears when people can raise risk, leaders can respond, and the decision changes the exposure in the field.

When culture scores rise but field decisions do not, leaders have found a transformation gap rather than a failed survey. Trace the decision, repair the ownership path, and verify the work again. That discipline turns listening into control.

Explore the Headline Podcast for more field-tested leadership perspectives.