How PepsiCo South America Cut Its Accident Ratio 50% in 6 Months and What the Metric Still Hid

PepsiCo South America cut its accident ratio by 50% in 6 months, and that result matters because a single number is only useful when leaders know what it hides. In 2024, the U.S. Bureau of Labor Statistics recorded 5,070 fatal work injuries, while the ILO continues to estimate nearly 3 million work-related deaths each year. Those figures are not decoration for a dashboard. They are the reason one green metric can never be the whole story.

This case study is written for executive leaders, dashboard owners, and EHS managers who need to know whether a better accident ratio reflects a safer system or only a cleaner number. Across 25+ years leading EHS in multinationals, Andreza Araujo has seen that the strongest metric question is not whether the score improved. It is what changed in the work that produced the score. In Safety Culture: From Theory to Practice, she argues that repeated decisions reveal culture. In The Illusion of Compliance, she shows why clean paperwork can still hide weak control.

The thesis of this article is direct. A 50% reduction in an accident ratio becomes useful only when leaders can explain the precursor signals, the field verification, the incentive design, and the decisions that made the improvement possible. Otherwise the company may have a better number while the same exposure logic survives in the next shift.

Initial scenario

The PepsiCo South America result came from a large, distributed operation, not from a single plant with one simple exposure. That matters because scale changes the meaning of the metric. A regional organization can improve a lagging number while still carrying different risk profiles across plants, warehouses, routes, contractors, and operating rhythms. In a network like that, the real question is whether the score moved because the system became safer or because the system became quieter.

ISO 45001:2018 already expects organizations to monitor, measure, analyze, and evaluate occupational health and safety performance. The standard, however, does not say that one headline metric proves control. That gap is where many dashboards fail. Leaders see a reduction, assume success, and stop checking whether field controls, work planning, and escalation rules changed in parallel.

Andreza Araujo has seen this pattern in more than 250 cultural transformation projects and across 30+ countries. Numbers improve faster when leaders change the questions they ask. The first question is no longer, "What is the rate?" It becomes, "What did we change that made the rate move?"

Decision

The key decision in any case like this is to stop using the accident ratio as the final proof of safety and start using it as one outcome in a larger evidence set. That is why the related article on Safety KPIs vs Bonuses vs Control Checks matters. A metric tied to reputation or compensation changes behavior in ways that may have nothing to do with risk reduction.

At PepsiCo South America, the useful decision was to pair the outcome number with leadership routines that looked at precursor data, field verification, and action quality. The organization did not need more decoration in the dashboard. It needed better decision rights around what the dashboard meant and who had to act when the signal changed.

That shift also fits James Reason's latent-condition lens. The visible result, whether good or bad, sits on top of earlier choices about planning, supervision, escalation, staffing, and control verification. If those choices do not change, the metric may improve for reasons that cannot be sustained.

Execution

The first execution step was to separate the lagging number from the operating evidence. A good accident ratio is still only one view, so leaders needed a second layer that could show whether the improvement was backed by field proof. That second layer included precursor signals, near-miss quality, corrective-action aging, and critical-control checks.

The second execution step was to make line leaders part of the metric conversation. A regional dashboard means little if only EHS knows how the number was created. The plant manager, supervisor, and site leadership team must understand which behaviors, checks, and escalations sit behind the score. Otherwise the metric becomes an executive artifact instead of a management tool.

The third execution step was cadence. A one-time review can make a number look impressive. A recurring review can expose whether the improvement holds. In practice, leaders needed a weekly line review, a monthly site review, and an executive review that asked different questions at each level. The wrong question at the wrong level is one of the fastest ways to turn a metric into theater.

For the board and senior team, the right companion to the accident ratio was not another vanity rate. It was a metric that answered whether the critical controls were actually being checked. The article on TRIR vs LTIFR vs DART vs SIF Rate explains why different metrics serve different decisions, and why a board should not ask a lagging number to do the work of a precursor signal.

Before and after

The table below shows the kind of shift that makes a 50% result meaningful. The number itself matters, but the operating change matters more because it tells leaders whether the gain can survive the next production push.

The value of that shift is practical. When leaders move from one number to four evidence streams, they reduce the risk of overconfidence. The organization still uses the accident ratio, but it no longer mistakes the ratio for the whole operating truth.

What the metric still hid

Even after a 50% reduction, the metric still hid several questions that matter to safety. Did the improvement come from better reporting quality, or from lower exposure? Did contractor work improve, or did only employee numbers improve? Did field verification become more disciplined, or did the team simply get better at closing visible actions before the review?

Those questions matter because a clean rate can coexist with underreporting pressure. Andreza Araujo's Portuguese title A Ilusao da Conformidade, glossed as The Illusion of Compliance, is useful here. Compliance can look elegant while the work still depends on shortcuts, silence, or luck. The related article on Metric Ownership shows why a metric without ownership becomes decoration.

This is also where the difference between TRIR, LTIFR, DART, and SIF rate matters. The article on which metric fits board decisions gives executives the comparison they need, because different metrics answer different questions. One number may be helpful for compliance, another for severity, and another for fatal-risk visibility, but none of them should stand alone.

Generalizable lessons

The first lesson is that a better accident ratio does not remove the need for precursor metrics. A 50% drop can be real and still incomplete. Leaders need at least one lagging indicator, one precursor indicator, one control indicator, and one response indicator so the dashboard can show history, exposure, verification, and action.

The second lesson is that ownership is part of the metric. If no manager can explain what the number means, the number is too far from the work. In more than 250 cultural transformation projects, Andreza Araujo has seen that metrics improve when line leaders understand them as operating cues, not as EHS ornaments.

The third lesson is that incentives shape truth. When the wrong number is tied to the wrong reward, the system starts protecting the number. That is why the article on Safety KPIs vs Bonuses vs Control Checks is a companion to this case. A dashboard that feels safe to the board but unsafe to report is not a good dashboard.

The fourth lesson is that a good result must be repeatable in 30 days, not only impressive in 6 months. If the same leadership team cannot explain how the score will be protected next month, the improvement may be vulnerable to turnover, schedule pressure, or contractor churn.

What to apply in your operation

Start with 4 metrics, not 40. Keep 1 lagging outcome, add 1 precursor signal, add 1 critical-control check, and add 1 management response metric. That mix gives a board enough information to govern and gives supervisors enough information to act.

Then test the metrics against 3 questions. Did the number change because exposure changed? Did the number change because reporting behavior changed? Did the number change because field verification changed? If the team cannot answer those questions, the metric is still too shallow for serious governance.

For the first 30 days, do not add a new bonus formula. Add a metric owner, a field verification sample, and a weekly review ritual instead. That sequence is usually enough to expose whether the number is a true signal or only a clean statistic.

Safety is about coming home, and that phrase only becomes operational when the dashboard helps leaders see what still threatens the journey home. If your organization needs help turning metric design into a safer management rhythm, *Safety Culture: From Theory to Practice* and Andreza Araujo's Safety Culture Diagnosis are the right next step.

Conclusion

PepsiCo South America's 50% accident-ratio reduction is useful because it shows that metrics can move when leadership behavior moves. The number is not the lesson by itself. The lesson is that a metric only becomes meaningful when executives can explain the precursor signals, the field checks, the ownership model, and the decisions that made the result possible.

For teams that want a dashboard that tells the truth, not just a dashboard that looks calm, start with Andreza Araujo's books, then test the operating system behind the number. The same company can have a better ratio and a weaker truth unless leaders keep asking what the metric still hid.