How a LATAM Food Operation Turned Behavior Signals Into a 50% Accident-Ratio Drop

The PepsiCo South America case is often remembered for one number: a 50 percent accident-ratio reduction in six months during Andreza Araujo's EHS leadership tenure. The safer lesson is more demanding. Behavior changed because leaders treated field signals as evidence of how work was really being done, not as a hunt for the person who needed another reminder.

On the Headline Podcast, Andreza Araujo and Dr. Megan Tranter often return to the gap between what leaders say about safety and what the operation rewards under pressure. This case belongs in that gap because behavior programs fail when they count observations while supervisors lack the authority, time, or discipline to change the condition that produced the behavior.

Why did behavior signals matter before the accident ratio moved?

Behavior signals mattered because lagging indicators could not show the daily choices that were making exposure normal. Before an accident ratio moves, supervisors usually see smaller evidence: a bypassed step, a rushed pre-task talk, a missing tool accepted as normal, a peer check skipped during peak demand, or a worker who stops raising the same concern because nobody closes the loop.

The market often treats behavior as the visible act at the end of the chain. James Reason's Swiss cheese model gives leaders a better frame because the visible act usually passes through planning, staffing, equipment, supervision, and culture before it reaches the worker. A behavior observation that does not inspect those layers is too shallow for serious prevention.

Andreza Araujo's book Safety Culture: From Theory to Practice argues that culture becomes visible in repeated decisions. In this case, the repeated decision was not only whether a worker followed a rule. It was whether each leader used the signal to adjust work before the next crew faced the same exposure.

1. Initial scenario

The initial scenario was a high-paced food operation where safety performance had to improve without pretending that production pressure would disappear. Food and beverage environments create tight cycles, frequent changeovers, contractor interfaces, logistics movement, sanitation pressure, maintenance windows, and line restart decisions whose timing affects how people behave.

A traditional behavior program could have responded by increasing observation quotas. More cards would have created more data, although more data alone would not explain why an operator skipped a check, why a supervisor accepted a workaround, or why a line team believed the safe method did not fit the shift rhythm.

The proprietary lesson from the PepsiCo South America case is that behavior had to be read as a diagnostic trail. The worker's action was treated as the beginning of the inquiry, while the end point was the leadership decision that changed the work condition.

2. Decision

The central decision was to stop treating behavior observations as an EHS policing routine. When observations become a quota, the observer looks for something to record and the worker learns to perform safety while being watched. That creates compliance theater, especially in work where the real exposure appears when nobody with a clipboard is present.

The better decision was to ask what each repeated behavior revealed about the work system. If people crossed a forklift route outside the marked path, the question was not only why they walked there. Leaders had to test whether the official path was too long, blocked, poorly marked, or unsupported by supervisor correction during shift peaks.

This is why the Headline article on closing a behavioral observation loop in 48 hours is an adjacent read. The loop matters because behavior data gains value only when the organization returns to the field with a decision, a correction, and proof that the exposure changed.

3. Execution

Execution required supervisors to convert signals into field checks. A behavior signal was not closed when someone spoke to the worker. It was closed when the supervisor checked the condition behind the behavior, named the owner, and verified whether the safe action had become easier or more reliable.

That execution discipline is different from telling people to be careful. A rushed lockout step may indicate weak planning, unclear isolation points, schedule pressure, or a restart habit that rewards speed. A missing peer check may indicate that the task is treated as routine, even though its energy, line-of-fire exposure, or changeover condition has changed.

The case also required leadership consistency. If one supervisor corrected the behavior while another supervisor praised the crew for recovering lost time through the same shortcut, the operation would teach two standards. Behavior signals become useful only when leaders remove mixed messages from the work.

4. Measured result

The measured result attached to this case is the 50 percent accident-ratio reduction in six months during the PepsiCo South America period. The important editorial boundary is that the number should not be stretched into a claim that every behavior program can produce the same reduction. Safety results depend on baseline risk, reporting quality, leadership authority, exposure mix, and operational discipline.

The result is still valuable because it shows that behavior work can become performance work when leaders act on signals. The reduction was not a motivational slogan. It was a sign that leadership routines, field conversations, and control verification had started to influence the conditions under which people made safety choices.

For YMYL safety writing, this distinction matters. A case study should preserve the verified data point and avoid false precision about mechanisms that are not publicly documented. The defensible lesson is that behavior signals must trigger operating decisions, not that one observation script fits every plant.

5. What changed in supervisor response?

Supervisor response changed from correction after the fact to diagnosis at the point of work. The supervisor still had to name the unsafe act when the risk was real, but the conversation could not end with blame, a signature, or a reminder. It had to identify which part of the work made the unsafe choice credible.

This connects directly with peer check, stop work, and pre-task briefing as behavior controls. Each tool works only when the supervisor protects the decision that the tool is supposed to create. A peer check without response authority becomes courtesy. A stop-work rule without managerial protection becomes a slogan.

A useful supervisor response had 4 parts: hear the signal, test the condition, hold the non-negotiable control, and close the loop. The order matters because a leader who starts with discipline may silence evidence, while a leader who only listens may leave the standard unclear.

6. Generalizable lessons

The first lesson is that behavior programs should measure response quality, not only observation quantity. A thousand observations can hide weak safety if most of them lead to coaching notes rather than control changes, work redesign, or resource decisions.

The second lesson is that repeated shortcuts are rarely random. Charles Duhigg's habit research helps explain why cues, routines, and rewards matter. If the cue is production delay and the reward is praise for recovery, the unsafe routine will survive a poster, a talk, or a training refresher.

The third lesson is that executives must review behavior signals as leading indicators. The executive dashboard should show repeated behavior themes, the condition behind each theme, the owner who changed the condition, and verification that the change reached normal work. Without that line of sight, leaders may celebrate a lower injury rate while the same behavioral pressure keeps rebuilding.

7. Before and after comparison

8. What should leaders apply in the next 30 days?

Leaders should choose one repeated behavior signal and run a 30-day response-quality test. Pick a signal that has appeared at least 3 times, such as missed peer checks, bypassed pedestrian routes, rushed pre-task reviews, incomplete handoffs, or repeated line-of-fire positioning. Then trace it from worker action to work condition to leadership decision.

The article on risk thermostat behavior shifts around controls helps explain why people adjust behavior when controls feel familiar, slow, symbolic, or easy to bypass. The 30-day test should therefore ask whether the control is practical under pressure, not only whether the worker knows the rule.

By the end of the month, the leadership team should have one field-verified change, one supervisor script that preserves the standard without blaming the worker, and one dashboard line that shows response time from signal to correction. That is a modest target, but it forces the behavior system to prove that it can change work.

Conclusion

The PepsiCo South America case shows that safe behavior improves when leaders convert signals into decisions, field checks, and control ownership, rather than treating observations as a count of worker actions.

Headline Podcast exists for leadership conversations about safety, culture, and the work that protects people. When your team needs a sharper discussion about behavior, pressure, and decision quality, follow Headline Podcast at headlinepodcast.us.