How a LATAM Food Operation Cut Accident Ratio 50% in 6 Months

A six-month accident-ratio reduction case shows how safety metrics become useful when leaders stop treating them as historical scorekeeping and start using them to govern exposure, follow-up quality, supervisor cadence, and field verification. The PepsiCo South America result matters because the measured reduction came from operating discipline, not from a prettier dashboard.

Many companies want the headline number without changing the measurement system that produced the old performance. They ask for fewer accidents, then keep the same monthly review, the same lagging dashboard, the same weak action follow-up, and the same habit of treating incident absence as proof that controls are working.

The stronger thesis is that the 50% reduction in accident ratio achieved during Andreza Araujo's PepsiCo South America tenure was not a metrics miracle. It was a management-system correction. The number moved because leadership changed what was reviewed, how fast weak signals were escalated, and which field evidence counted as proof of control.

Initial scenario

The initial scenario was familiar to any large food operation spread across multiple countries. Sites had different maturity levels, different supervisor routines, different interpretations of risk, and different ways of translating corporate safety language into field action. A central dashboard could show accident counts, but it could not, by itself, explain why a warehouse, factory, or distribution center was drifting toward harm.

During her role as EHS and Sustainability Manager for LATAM, Caribbean, and Central America at PepsiCo Foods, Andreza Araujo operated across 7 countries, 30 factories, and 168 distribution centers. That scale matters because a metric that works in one plant can become noise when it is copied across a regional operation without clear definitions, follow-up rules, and leadership cadence.

The core problem was not that lagging indicators were useless. The problem was that lagging indicators were arriving too late and receiving too much authority. The related article on lagging indicator limits explains the same trap: a low number can describe the past while saying little about tomorrow's exposure.

Decision

The critical decision was to stop reading accident ratio as an isolated outcome and start treating it as the visible end of a chain. If the chain included weak supervisor routines, slow corrective actions, poor hazard recognition, inconsistent observation quality, and unclear escalation, then the accident ratio would only improve when those upstream elements were managed with the same seriousness as the final number.

Across 25+ years leading EHS at multinationals, Andreza Araujo has identified one pattern repeatedly: senior leaders often ask for a better safety result while leaving the operating rhythm untouched. That creates pressure to explain the number, not capability to change the exposure that creates the number.

As Andreza Araujo argues in Safety Culture: From Theory to Practice, culture becomes visible through what leaders reinforce and tolerate. In a metric system, that means the review agenda must reward verified control, not only a quiet month. When the meeting celebrates a low injury rate without asking what was found, fixed, and learned, the organization trains people to protect the number.

Execution

The execution phase needed a different measurement rhythm. The regional operation had to connect accident ratio to leading evidence that supervisors and managers could influence every week. That evidence included action closure quality, observation quality, recurring unsafe conditions, training gaps tied to real tasks, and the speed with which sites escalated repeated weak signals.

This is where many safety metric programs fail. They add more indicators, but they do not define the management response attached to each one. A metric without a required decision becomes decoration. The article on safety KPIs, bonuses, and control checks shows why the behavioral incentive behind a KPI matters as much as the calculation.

The stronger operating question was not whether the dashboard looked complete. It was whether the dashboard forced the right conversation before serious harm occurred. If a distribution center had repeated vehicle-pedestrian conflicts, the review had to ask whether route separation, supervision, contractor control, and corrective action aging were moving, not whether the month had stayed injury-free.

Measured result

The measured outcome was a 50% reduction in accident ratio in six months during Andreza Araujo's PepsiCo South America tenure. That figure should be read carefully. It is not a universal promise that any company can halve accidents in the same window. It is evidence that accident reduction becomes plausible when the leadership system, the metric system, and field execution finally point in the same direction.

The result also shows why an executive dashboard must separate performance history from control confidence. Accident ratio tells leaders that harm has changed. It does not automatically prove which controls improved, which risks remain hidden, or which sites are underreporting weak signals. The article on SIF rate, TRIR, and precursor metrics develops that distinction for serious injury and fatality prevention.

What changed in the metric system

The most important change was the loss of innocence around the number. A low accident ratio stopped being treated as a sufficient story. Leaders had to ask what the number was hiding, which controls had been verified, and whether workers were reporting weak signals before injuries appeared.

That shift is especially important in food manufacturing and distribution, where routine work can normalize exposure quickly. Forklift interaction, machine guarding, manual handling, sanitation chemicals, line speed, contractors, and maintenance interventions can all create risk that will not appear in the injury rate until the system has already tolerated too much drift.

In more than 250 cultural transformation projects supported by Andreza Araujo's team, one repeated lesson is that measurement changes behavior only when it changes leadership attention. If the review still spends 80% of its time on red and green injury charts, the new indicators will not matter. If the review asks which repeated condition was removed, which supervisor routine improved, and which control was verified, the metric begins to govern reality.

Generalizable lessons

The first lesson is that accident reduction should not be delegated to the EHS department. EHS can design the metric architecture, but operations owns the conditions that create exposure. When supervisors and plant leaders do not own the weekly evidence, the regional number becomes a report about work rather than a control over work.

The second lesson is that faster follow-up is not the same as better follow-up. Closing actions quickly helps only when closure changes the condition, removes exposure, improves supervision, or strengthens a barrier. The article on corrective action aging dashboards is useful because aging reveals where management attention is losing contact with risk.

The third lesson is that safety metrics must be translated into local decisions. A regional team may define the standard, but the factory manager, distribution leader, and supervisor need to know what to do on Monday morning when a leading indicator turns red. Without that translation, the dashboard becomes a regional artifact rather than an operating tool.

What to apply in your operation

Start by separating 3 families of indicators: harm outcomes, exposure signals, and control evidence. Harm outcomes include accident ratio, LTIFR, DART, severity rate, and recordable events. Exposure signals include repeated unsafe conditions, task-risk drift, near misses, contractor deviations, and recurring ergonomic or traffic conflicts. Control evidence includes completed field verification, quality of corrective actions, supervisor routines, and critical-control checks.

Then define what decision each indicator triggers. A repeated exposure signal should trigger a field review, not only a comment in the monthly meeting. A corrective action that ages beyond the agreed threshold should trigger escalation, not a reminder email. A site with no reported near misses for months should trigger a reporting-quality conversation, because silence is not the same as safety.

For teams that need a clean operating language, a safety metric dictionary can prevent one of the most common failures in multi-site operations: each site using the same word for different realities. Definition discipline is not bureaucracy. It is what allows leaders to compare risk honestly.

Where leaders should be careful

The first trap is copying the 50% figure as a target. When a result becomes a quota, leaders may create pressure to suppress reporting or reclassify events. The better lesson from the PepsiCo South America case is not the percentage alone, but the operating discipline that made the percentage credible.

The second trap is adding leading indicators without reducing meeting clutter. If every measure is presented with the same weight, leaders will return to the number they understand fastest. Keep the dashboard small enough for decisions and strict enough to expose whether controls are actually changing.

The third trap is using metrics to judge sites before understanding their reporting culture. A site with more reported weak signals may be more honest, not more dangerous. Andreza Araujo's Portuguese title Muito Alem do Zero, glossed as Far Beyond Zero, is relevant here because zero-accident language can make leaders confuse silence with excellence.

The PepsiCo South America result is valuable because it refuses a comfortable interpretation. Accident ratio did not fall because leaders admired a better number. It fell because the organization changed the way safety performance was governed across countries, factories, and distribution centers.

Visit andrezaaraujo.com to explore Andreza Araujo's books, diagnostics, and corporate programs for companies that need safety metrics to reveal operational truth before harm becomes visible.