New Safety Data Steward in 45 Days: Turn Reporting Rules Into Trusted Decisions

A new safety data steward can inherit a dashboard that looks precise while the decisions behind it remain vague. The first 45 days should not be spent polishing charts. They should establish which data describes exposure, which evidence supports a control claim, who owns the response, and when a trend requires escalation.

The role sits between operations, EHS, maintenance, HR, and leadership. That position creates a practical risk. If the steward treats every number as equally reliable, the organization may reward clean reporting while missing changing conditions in the field.

What should a new safety data steward understand before starting?

A safety data steward is responsible for the meaning, quality, ownership, and use of safety information. The role is not limited to entering incident records or maintaining a dashboard. It makes sure that leaders can distinguish an event count from an exposure signal, a completed action from a verified control, and a missing record from an absence of risk.

ISO 45001:2018 requires organizations to monitor, measure, analyze, and evaluate performance, yet a management system cannot make a weak data definition reliable. The steward has to connect each measure to a decision, a data owner, a collection method, and a review cadence.

As Andreza Araujo argues in Safety Culture: From Theory to Practice, safety culture becomes visible through everyday decisions. Data is part of that culture because people learn what leaders value from the questions asked after a number changes.

Days 1 to 7: Map what each number is supposed to mean

Start by listing every safety measure that appears in leadership meetings, site reviews, shift boards, and monthly reports. Include injury rates, near-miss counts, inspection completion, critical-control verification, overdue actions, training completion, worker concerns, and exposure measures when the operation uses them.

For each measure, write one sentence that answers four questions. What decision should this number support? Who owns the source record? How is the value calculated? What would make the number misleading? A measure without a decision purpose is usually a habit, not an indicator.

Compare the dashboard with the records that generate it. If a safety dashboard presents a green status, trace the status to its source and ask whether the evidence describes work as performed rather than work as planned. The existing guide on safety dashboard blind spots provides a useful contrast, while your first task is to document how those blind spots could appear locally.

By day 7, create a measure register with five fields: measure name, definition, owner, source, and decision use. Add a sixth field for known limitations. The limitation matters because a transparent weakness is safer than a precise-looking number that no one can challenge.

Days 8 to 15: Separate reporting rules from decision rules

During the second phase, review the rules that govern when people report an event, classify a case, close an action, or escalate a concern. Reporting rules determine what enters the system. Decision rules determine what leaders do with it. They are related, but they are not interchangeable.

A near miss may increase because reporting became easier, because exposure increased, or because both changed at the same time. A lower count may indicate improvement, reduced reporting, fewer opportunities to observe, or a classification change. The steward should never explain a trend before checking which of these conditions is plausible.

Use the article on leading safety indicators to challenge the assumption that every preventive measure proves prevention. For each leading measure, record what it can show, what it cannot show, and what additional evidence a leader needs before changing the work.

Set a decision rule for data quality. If the source is incomplete, late, duplicated, or classified inconsistently, the dashboard should show that limitation instead of silently filling the gap. A warning about data quality is not an administrative failure. It is evidence that the organization is protecting the decision from false confidence.

Days 16 to 23: Test the data at the workface

In the third phase, follow three measures from the report to the worksite. Choose one event-based measure, one control or action measure, and one worker-input measure. Ask the people who create the records what happens when conditions change, when a supervisor disagrees with a classification, and when an issue has no obvious owner.

Compare the written record with what the crew understands. A completed inspection may show that a form was submitted, while the field may reveal that the inspection did not test the control that mattered. An open concern may appear overdue because the response was not recorded, or it may be overdue because the assigned owner lacked authority to change the exposure.

Andreza Araujo's work across more than 250 cultural-transformation projects supports a practical principle: the value of a safety number depends on the decision it improves close to the work. The steward should therefore test data definitions with operators, supervisors, and maintenance personnel, whose experience often reveals ambiguity earlier than a monthly review does.

By day 23, document at least three examples in which the source record and the field condition differ, then classify the reason. The gap may come from timing, ownership, access, classification, or a control that was never verified. Do not turn this exercise into blame. Its purpose is to repair the path between observation and decision.

Days 24 to 30: Build a trustworthy review conversation

A trusted review does not begin with a chart. It begins with the operational question that the chart is meant to answer. Ask whether exposure changed, whether a control remains reliable, whether an action changed the work, or whether leaders are receiving the same concern repeatedly without removing its cause.

Give every recurring measure a standard discussion sequence. State the change, identify the evidence, name the uncertainty, assign the decision owner, and record the next verification. This structure keeps a meeting from ending with commentary but no accountable response.

Link the review to a short decision checkpoint, such as the routine described in this 20-minute safety decision checkpoint. The checkpoint should not become another meeting that copies the dashboard. It should be the point at which evidence changes a work plan, a control, a resource decision, or an escalation.

Invite disagreement before publishing the monthly version. A supervisor may know that a measure is inflated by duplicate entries. A technician may know that a control is recorded as present even when access conditions make it unreliable. A worker may know that concerns disappear when they are routed through a system no one owns.

Days 31 to 37: Set ownership and escalation thresholds

During this phase, assign an owner to each material data problem. The data steward owns the integrity of the information flow, but the operational leader owns the condition that requires action. Keeping those responsibilities separate prevents EHS from becoming the permanent owner of risks it cannot directly change.

Define thresholds before the next difficult case arrives. Escalation may be triggered by a repeated control failure, a high-consequence exposure without evidence, a growing delay in corrective action, a worker concern that returns after closure, or a sudden change in reporting volume that lacks an explanation.

Use the article on safety decision latency to examine how long information takes to become a decision. A threshold should include a response owner, an expected response time, and the evidence that confirms the response worked. Without those elements, escalation becomes a message rather than a control.

During her PepsiCo South America Foods tenure, Andreza Araujo led a 50% reduction in accident ratio in six months through a defined transformation plan. The transferable lesson is not a promise that every operation can reproduce the same result. It is that measurement becomes useful when it is connected to a deliberate operating plan and visible leadership action.

Days 38 to 45: Pressure-test the system before handing it over

In the final phase, run a controlled review using a real change in conditions. Choose a delayed maintenance activity, a staffing change, a contractor interface, a production increase, or another situation in which the data flow is likely to be stressed. Ask the team to show what the dashboard would report, what it would omit, and who would decide whether work continues.

Test the system with an intentionally incomplete record. The objective is not to create confusion. It is to see whether the organization marks uncertainty, seeks the missing evidence, and escalates when the exposure cannot be bounded. A system that displays a confident green status despite missing evidence is not ready for handover.

At day 45, deliver a short stewardship report. Include the five measures that leaders can trust, the three measures that need repair, the decisions each measure should support, the owners of unresolved data gaps, and the escalation thresholds that are now active.

Close the transition with the operations leader, not only with the analytics or EHS team. The handover is complete when leaders agree which numbers deserve attention, which limitations must remain visible, and which field evidence will change the next decision.

Common mistakes that weaken a safety data steward's first 45 days

The first mistake is treating completeness as quality. A fully populated form can still describe the wrong exposure, use an unclear definition, or record a control that no one verified.

The second mistake is changing every measure at once. A new steward creates confusion when definitions, dashboards, ownership, and meeting routines all change before anyone understands the existing failure points. Stabilize the meaning of a small set of important measures before expanding the system.

The third mistake is becoming the owner of every response. The steward should expose the decision and protect the evidence, while the leader with authority changes the work, funds the control, adjusts the plan, or stops the activity.

The fourth mistake is hiding uncertainty to protect credibility. Credibility grows when the steward can say what is known, what is not known, and what evidence will close the gap. This is especially important when leadership wants a simple status from a complex operation.

Resources to deepen the role

Make The Difference: Be a Leader in Health & Safety is useful for connecting data review with daily leadership behavior. Safety Culture Diagnosis: Learn how to do your own helps practitioners examine how people experience the systems that produce safety information. Across 25+ years in multinational EHS, Andreza Araujo has kept the same test in view: safety information should help people make better decisions before someone is harmed.

A safety data steward does not make an operation safer by producing more reports. The role creates value when definitions are clear, limitations are visible, evidence reaches the workface, and the person with authority acts before uncertainty becomes exposure. Safety is about coming home, and trusted data should help leaders protect that outcome.

If your organization needs help connecting safety data, culture, and operational decisions, explore Andreza Araujo's safety culture work.