How 2,300 Behavior-Based Safety Projects Revealed the Leadership Variable That Changes Field Behavior

A five-year review of 2,300 behavior-based safety projects reached a conclusion that changes where leaders should look first. According to Dr. Thomas Krause on the Headline Podcast, the strongest predictor of success was the quality of leadership given to the initiative, not simply whether the program reached frontline workers.

That finding matters because many programs begin with the visible part of the intervention. The organization trains observers, prints cards, sets a monthly target, and asks supervisors to increase conversations. The harder question comes later. What does a leader do when the conversation reveals a production constraint, a weak control, or a shortcut that the system quietly rewards?

Initial scenario

A behavior program usually enters an operation as a promise to improve the next decision at the point of work. It fails when the organization measures participation while leaving the conditions around that decision untouched. The initial scenario is therefore not a careless worker. It is a useful intervention entering a system that has not decided how leaders will respond.

The distinction is important for supervisors. A worker may place a hand inside a line of fire because the approved tool is unavailable, because the task changed after the briefing, or because the normal method adds a delay that the shift has learned to absorb. The observation records the action, but the leadership response determines whether the organization sees a person to correct or a control to repair.

Dr. Thomas Krause's finding gives this problem a measurable frame. A program can have high participation and still be weak if the leadership surrounding it is inconsistent. When one supervisor asks curious questions and another opens with blame, the organization is not running one intervention. It is running several local systems with different consequences for telling the truth.

Andreza Araujo's work makes the same boundary visible from a culture perspective. In *A Ilusão da Conformidade*, psychological safety is treated as the ground where prevention can function because risk information does not circulate when people expect exposure or punishment. The behavior conversation is therefore a leadership test before it is a worker test.

Decision

The decision was to stop treating observation volume as the primary proof of progress and to examine the leadership conditions that determine whether each conversation changes the work. That means asking what the supervisor noticed, what response was made possible, who owned the next action, and how the operation verified the result.

This is not a rejection of observation. Observation gives leaders access to the task as it is performed, which is valuable evidence when it is handled with discipline. The decision is to move the program's center of gravity from collection to response. A card that travels into a database but never changes a tool, schedule, barrier, or decision right is activity, not improvement.

The leadership variable also changes the supervisor's role. The supervisor is not a messenger who repeats a rule after a deviation. The supervisor is the person who can connect the worker's explanation to the conditions of the job, escalate an unresolved conflict, and protect time for a safer method when the shift is under pressure.

That is why a leader's first question matters. A question such as “What made this choice seem reasonable?” can expose a missing guard, an ambiguous standard, or a workload problem that a corrective lecture would hide. Curiosity does not remove accountability. It produces better evidence before accountability is assigned.

Execution

Execution begins with a small set of leadership behaviors that can be observed in the field. Leaders model the conversation, remove practical barriers, close the response loop, and review whether the change survives the next shift. The program becomes credible when workers can see that speaking honestly produces a better decision rather than a faster reprimand.

The first move is to model the conversation. Dr. Thomas Krause described walking the floor with leaders and asking what people like about the job, what makes them nervous, and what could be better. The value is not the script itself. The value is that the leader learns to listen before defending the existing arrangement.

The second move is to separate a behavior observation from a behavior verdict. The observation should describe the action, context, exposure, and available controls. A verdict jumps directly to character. Once the record says that a worker was careless, the team has little reason to add the operational detail that could prevent recurrence.

The third move is to assign the response to the department that runs the work. The safety function can help define the risk and test the control, but it should not become the permanent owner of every fix. A change that belongs to maintenance, engineering, planning, or operations will not last if those functions only receive a copied action from EHS.

The fourth move is to make the response visible. A worker who raises a concern should learn what was changed, what was not changed, who owns the decision, and when the result will be checked. Silence after a report teaches the workforce that participation is a one-way transaction.

These moves are practical, but they require leadership capacity. If supervisors have no time to observe the work, no authority to pause a task, and no route to resolve a design problem, the program will eventually reduce itself to conversation quality theater. The intervention must therefore be connected to the operating rhythm that controls resources and priorities.

Measured result

The documented result is not a promise that every project produced the same outcome. The review's usable result is that leadership quality predicted success across 2,300 projects tracked for five years. The operational implication is to measure the leadership response around each observation, because that is the mechanism that separates a durable change from a completed form.

The table below translates the finding into indicators a site can verify without pretending that activity equals safety. It distinguishes the old signal, which mainly counts participation, from the stronger evidence that shows whether leadership has made safer behavior more likely.

These indicators also protect the program from a familiar distortion. When a target is attached to the number of observations, people can optimize the count by recording easy, low-consequence behaviors while critical exposures remain unspoken. A smaller set of high-quality conversations can create more value than a larger set of forms that never reach a decision owner.

The strongest result is therefore not a green dashboard. It is a visible change in the next task, supported by a leader who can explain what changed and why. Where a control cannot be improved immediately, the leader should document the residual risk, the temporary protection, and the escalation path rather than closing the record with a generic reminder.

Generalizable lessons

The 2,300-project finding generalizes through the mechanism of leadership response. The same observation process can produce learning in one operation and silence in another because the surrounding leaders create different consequences for evidence, challenge, and follow-through.

1. Leadership quality is visible in the second question

The first question often confirms that a rule exists. The second question reveals whether the leader wants to understand the work. Asking what made the choice sensible at the time opens the path to workload, tool availability, design, and competing goals.

2. A conversation needs a decision route

Workers stop contributing when concerns disappear into a system. Every meaningful observation needs a route to a person who can decide, fund, redesign, or escalate the response. The route should be clear before the campaign begins.

3. Coaching cannot replace control improvement

A respectful conversation is not the final control. If the same exposure returns, the organization has learned what to say without learning what to change. Coaching is useful when it leads to a better control, clearer authorization, or safer task design.

4. The field is the verification point

Minutes, action logs, and completion percentages are supporting evidence. The decisive test happens where the task is performed again. A leader should return after the change and ask whether the new method is usable under the same pressure that produced the original behavior.

For a practical companion, compare this case with the behavioral observation conversation that changes the next task and the choice between coaching, observation, and task authorization.

What to apply in your operation

Choose one high-consequence exposure and follow the full response loop for thirty days. Observe the work, ask what made the current choice reasonable, assign the operational owner, verify the control in the field, and report what happened back to the people who raised the concern.

Start with a single area rather than launching another enterprise-wide campaign. Select a task where the gap between the written method and the performed method is already visible, then involve the supervisor who controls the daily conditions. The goal is not to stage a perfect conversation. It is to make the next decision more reliable.

At the end of the thirty days, review five questions. Did the leader spend time at the point of work? Did workers add information after the first answer? Did the department that runs the task own the change? Did the control survive a busy shift? Can the leader show the evidence without relying on observation counts alone?

If the answer is no, do not rescue the program with more posters or a larger quota. Repair the leadership condition that blocked the response. As Andreza Araujo argues through the culture principles in *Liderança Antifrágil*, prevention strengthens when leaders make bad news usable and turn it into a better decision before the next event.

For a second lens, read the risk perception distortions that change a safe decision. The behavior program becomes credible when the organization connects what people perceive, what leaders permit, and what the operation actually controls.

The lesson from 2,300 projects is direct. Field behavior changes when leadership changes the conditions around the field decision. Count the conversation if you need a starting signal, but judge the program by the control, decision, or work method that becomes safer afterward.