How a Metric Reset Made a 50% Accident Drop Credible in Six Months

The PepsiCo South America case is often quoted for one number, the 50% accident ratio drop in six months during Andreza Araujo's EHS leadership tenure. That number matters, but only if the metric system around it stayed honest. A lower count can mean less exposure, or it can mean quieter reporting, a confused denominator, or a leadership rhythm that rewards tidy dashboards over useful evidence.

A metric reset in safety is the point at which leaders stop treating injury counts as proof and start testing whether denominators, reporting quality, action aging, and field verification still describe reality.

That distinction is the reason this case belongs in safety indicators and metrics. As Andreza Araujo writes in The Illusion of Compliance, the real system is what happens when no one is watching, not what the spreadsheet says after the week closes.

Initial scenario

A consumer goods region can look healthy on paper long before the field feels healthy. Monthly reports improve, charts turn greener, and the same leadership team begins to believe the operation is moving in the right direction. Yet the real question is not whether the line moved. The question is whether people still report the weak signals that prevent the next serious event.

Across 25+ years of executive EHS leadership and more than 250 cultural transformation projects, Andreza Araujo has seen the same pattern repeat. Leaders celebrate a good line graph before they ask whether the workforce still feels safe enough to surface bad news, whether the denominator still matches exposure, and whether the metric is describing the work or merely the paperwork.

The first trap is simple. A dashboard that gets greener while the field gets quieter is not a success signal. It is a signal to test whether the organization has become more capable or merely more selective in what it records. That is exactly why this article sits in the safety indicators and metrics category rather than in a generic safety culture bucket.

Decision

During Andreza Araujo's PepsiCo South America tenure, the accident ratio fell 50% in six months. The key decision was to treat that number as a management test, not as a trophy. A result that fast deserves respect, but it also deserves skepticism until the organization proves that the metric still reflects reality.

The team therefore looked at four things together, injury frequency, near-miss quality, action aging, and control verification. James Reason helps here because latent failures rarely announce themselves in the final number. They sit earlier, in planning gaps, supervisor drift, maintenance backlog, and weak response to concerns that never received enough force.

As Andreza Araujo argues in Safety Culture: From Theory to Practice, culture appears in repeated decisions. In this case, the repeated decision had to be that a report would trigger action, not punishment, and that a control would be verified in the field before the metric was allowed to improve on paper.

Execution

The execution step was not to pile on more data. It was to make the existing data harder to fake and easier to use. A metric reset works only when leaders agree on what each indicator means, which base it uses, who owns the response, and how fast the field will know whether the number changed for a real reason.

That is where denominators matter. If one site counts headcount, another counts hours, and a third counts task volume, the board sees three different stories and calls them one dashboard. The article on safety metric denominators explains why exposure base choice changes the meaning of a rate, while underreporting risk shows how a cleaner line can still hide a weaker system.

The practical move is to pair lagging and leading signals. Injury ratio is only one piece. Near-miss quality, action aging, field verification, and response latency show whether the organization is learning or just counting. In more than 250 cultural transformation projects, Andreza Araujo has repeatedly seen that the leaders who move fastest are the ones who change the operating rhythm, not the ones who ask for a prettier report.

That is also where the metric vocabulary should change. If the team can say that a report was raised, verified, acted on, and closed in the field, the metric tells a stronger story than a simple count. If the team can only say that the issue is still being watched, the metric is probably documenting attention rather than control.

Measured result

The measured result was a 50% accident ratio drop in six months. That figure is meaningful because it is tied to a verified period in Andreza Araujo's EHS leadership biography, but it remains useful only if the reporting system stayed credible while the number moved down.

One way to test credibility is to ask whether near-miss quality rose while the injury ratio fell. Another is to ask whether action aging shortened rather than merely shifting the problem into the next quarter. If both improved, the metric reset was probably describing reality. If only the injury line improved, leaders should assume the picture may be incomplete.

The market often wants a clean number and a simple story. This case refuses that shortcut. A lower count may be good news, but it is not proof by itself. The organization still has to show that the denser control network, the faster response, and the better field checks match the number that got better on the slide.

Generalizable lessons

The first lesson is that a good metric can still be the wrong metric if the base is mismatched. Injury counts without exposure context invite false comfort, because a smaller number can reflect less work, less reporting, or less risk. The board needs to know which of those three changed before it assumes the improvement is real.

The second lesson is that reporting quality is part of safety performance, not a side issue for communications. If people stop reporting because the number became politically sensitive, the dashboard improves while the system loses sight of the real hazard. That is why Andreza Araujo's book The Illusion of Compliance is useful here. It reminds leaders that a tidy file can coexist with a weak operating reality.

The third lesson is that action aging belongs beside the injury rate. A metric that closes cleanly in the spreadsheet but ages badly in the field teaches people that delay is acceptable. Once that lesson lands, the score may look better while the exposure continues to wait.

The fourth lesson is that field verification matters more than presentation polish. If the control is only confirmed in a meeting, the metric describes confidence. If the control is confirmed in the work area, the metric starts to describe capability.

What to apply in your operation

For an EHS manager, the next move is to separate the dashboard into three questions. What changed in exposure, what changed in reporting, and what changed in control quality. If those three answers are not distinct, the numbers are still too blended to guide action.

For a supervisor, the practical job is to see whether the field still behaves the way the dashboard claims. If the ratio improved but the crew still skips a verification, the supervisor should not congratulate the number. The supervisor should correct the condition that let the shortcut survive.

For an executive, the useful question is whether the organization can explain the result without leaning on volume rhetoric. A result that depends on a single celebrated number is fragile. A result that can be traced to protected reporting, faster action, and verified controls is much harder to fake.

For a risk or compliance lead, the reset means reviewing whether the metric dictionary is still fit for purpose. The definitions must be specific enough that the same event is counted the same way every time, because consistency is what lets leaders compare periods without fooling themselves.

That is the practical value of the case. It turns a result into a method, which is the only form of case study that survives beyond applause.

Before and after metric read

The uncomfortable part is that a better number does not guarantee better control. It only means the leadership team has one more thing to verify before it calls the result real.

FAQ

Is a 50% drop enough to prove improvement? No. It is a strong signal, but it still needs denominator checks, reporting quality checks, and field verification before leaders treat it as proof.

Which metric should leaders add first? Action aging is usually the fastest upgrade, because it shows whether unresolved risk is being carried forward instead of closed in the field.

How do you know if the denominator is wrong? When different teams count the same risk differently, or when the metric base no longer reflects exposure hours, task volume, or the real unit of work, the denominator is not helping leaders compare reality.

What should a supervisor watch in the field? The supervisor should watch whether the control is actually being used, whether the crew can explain it, and whether the response to a weak signal is quick enough to change the next shift.

How does this connect to Andreza Araujo's books? The Illusion of Compliance explains why tidy reporting can mislead, while Safety Culture: From Theory to Practice explains why repeated decisions create the real operating culture.

Headline Podcast exists for this exact kind of conversation. If your dashboard looks better than your field, share the case with the person who owns metrics, not only the person who owns the slide. Then ask whether the number improved because the work changed or because the story got cleaner.

Listen to Headline Podcast