How to Audit Leading Indicator Quality in 30 Days

A leading indicator can protect people only when it shows whether a control is working before harm appears. Many safety dashboards do something weaker. They count inspections, observations, talks, audits, training hours and open actions, then present the volume as prevention. The number rises, the dashboard looks active, and the operation still does not know whether fatal exposure has changed.

That is why EHS managers need a leading indicator quality audit, not another dashboard refresh. OSHA's publication Using Leading Indicators to Improve Safety and Health Outcomes describes leading indicators as proactive, preventive and predictive measures that reveal potential problems in the safety and health program. ANSI/ASSP Z16.1-2022 also pushes a balanced metric system that connects leading, lagging and impact measures. The practical test is blunt: if the indicator moves, does anyone know what decision should change?

This guide uses a 30-day audit cycle for EHS managers, safety analysts and site leaders who already track leading indicators but no longer trust the signal. It complements the existing article on leading indicators TRIR will never show by focusing on metric quality rather than metric selection.

A leading indicator quality audit is a structured review of proactive safety metrics to verify whether each indicator has a clear definition, accountable owner, reliable data source, risk connection, field evidence and decision rule. Its purpose is to remove dashboard noise before leaders mistake activity for prevention.

What you need before starting

Start with a narrow sample, because a full dashboard audit can become bureaucratic before it becomes useful. Select 8 to 12 indicators that leaders already use or claim to use. Include at least one indicator tied to serious injury and fatality exposure, one tied to contractor risk, one tied to corrective actions, one tied to safety conversations, and one tied to verification of critical controls.

You also need access to the raw records behind the dashboard. A number presented as 96 percent complete says little until the auditor sees what counts as complete, who entered the data, what evidence was required, and whether a supervisor could close the item without changing the field condition. The companion article on metric hygiene and data defects gives the data-quality lens for this part of the work.

Across 25+ years in executive EHS and more than 250 cultural transformation projects, Andreza Araujo has seen that indicator systems often fail less from lack of data than from weak interpretation. In A Ilusão da Conformidade, glossed in English as The Illusion of Compliance, the point is that good indicators do not guarantee good practices. The audit exists because that sentence is operationally true.

Step 1: Freeze the indicator list for 30 days

The first step is to freeze the audit scope so the team does not keep adding indicators whenever discomfort appears. Pick the indicators, define the site or business unit covered, and state the 30-day period under review. If leaders want a global dashboard audit, use one site as the pilot and expand only after the method works.

For each indicator, record the current title, definition, owner, formula, data source, reporting frequency, target, and audience. Do not rewrite anything yet. The first audit file should capture the metric as it actually operates, because hidden ambiguity is part of the evidence.

The common error is starting with design preference. Teams argue about what the dashboard should look like before they understand whether the existing number can be trusted. A useful audit starts colder than that, with the current rule, current record and current decision path.

Step 2: Test whether each definition can be applied by two people

The second step is a reproducibility check. Give two competent people the same 10 records and ask them to classify whether each record counts for the indicator. If they disagree often, the indicator is not yet a metric. It is an opinion with a percentage sign attached.

This matters for observations, safety conversations, hazard reports and action closures. One supervisor may count a short comment as a safety conversation, while another requires a documented risk discussion with agreed action. Both may be acting in good faith, but the dashboard cannot compare their results if the definition changes by person.

Use a simple pass rule. If two reviewers cannot reach at least 90 percent agreement after reading the definition, rewrite the definition before the indicator is used for ranking, bonuses or executive comparison. The point is not statistical perfection. The point is to prevent leadership from making decisions on a measure whose counting rule changes by site.

Step 3: Trace each indicator to a risk control

The third step is to trace each indicator to the risk control it is supposed to strengthen. ISO 45001:2018 clause 9.1 requires organizations to monitor, measure, analyze and evaluate occupational health and safety performance. That requirement becomes meaningful when the measure is connected to a control, not just to an activity.

Ask one question for every indicator: which exposure would get worse if this number deteriorated? If the answer is vague, the metric is not ready. A toolbox-talk completion rate may be relevant for communication discipline, but it does not prove machine guarding, line-of-fire control or confined-space isolation unless the content, audience and field verification link to those controls.

This is where the audit should connect to SIF rate, TRIR and precursor logic. A leading indicator that cannot be traced to a serious-risk precursor may still be useful, but it should not be sold as fatality prevention.

Step 4: Check the owner behind the number

The fourth step is ownership testing. Every indicator needs a data owner, a process owner and a decision owner. In smaller sites, one person may hold more than one role, but the roles still need names. Otherwise, the indicator becomes a shared orphan that everyone reads and nobody improves.

The data owner ensures the record is entered correctly. The process owner improves the underlying work, such as inspection quality or action closure. The decision owner changes resources, priorities, escalation or supervision when the indicator shows deterioration. If any of these roles is missing, the metric may continue to report risk while the organization continues to tolerate it.

In more than 250 projects supported by Andreza Araujo, weak ownership often appears as polite dashboard language. The team says that the metric is monitored by EHS, reviewed by operations and discussed in meetings, but nobody can name the person authorized to change the condition that the metric reveals.

Step 5: Sample the raw evidence behind the dashboard

The fifth step is evidence sampling. Pull 10 to 20 records for each indicator, depending on volume, and verify whether the record proves what the dashboard claims. For completed inspections, look at the inspection form, the finding quality, the photo evidence where available, and the action generated. For closed actions, check whether the field condition changed.

Do not sample only clean records. Include late records, perfect records, records entered near the deadline, records from high-performing sites, and records from sites that report unusually low exposure. The audit is looking for distortion, and distortion often hides in results that look too smooth.

A common trap is treating system timestamps as proof of control. A record closed at 4:58 p.m. before the monthly dashboard cut-off proves that someone clicked a system field. It does not prove that the guard was repaired, the chemical container was relabeled, or the contractor changed the method statement.

Step 6: Compare dashboard signal with field reality

The sixth step is field verification. Pick a small number of locations where the indicator says performance is strong and verify whether the worksite confirms the signal. If pre-task risk assessments show 100 percent completion, visit the area and ask workers what changed in the task because of the assessment. If safety observations increased, check whether repeated exposure declined.

This is not a hunt for errors. It is a validity check. The article on indicator triangulation and cross-checks for risk explains why one metric should be compared with field evidence, control checks and lagging outcomes before leadership treats it as reliable.

The strongest audit question is simple enough for a supervisor to answer. What did this number help you see earlier than an incident would have shown? If the supervisor cannot answer, the indicator may be a reporting habit rather than a prevention tool.

Step 7: Define a decision rule for every retained indicator

The seventh step is to attach a decision rule to each indicator that survives the quality checks. A decision rule defines what happens when the metric crosses a threshold, deteriorates for two cycles, improves unexpectedly, or conflicts with another signal. Without that rule, the dashboard can be reviewed forever without governing anything.

Examples are practical. If critical-control verification quality drops below the agreed threshold for two weeks, the plant manager joins the field verification walk. If corrective-action aging exceeds the limit, the owner must separate resource constraints from low-priority actions. If high-potential near-miss reporting falls while overtime rises, the site must test whether reporting trust has weakened.

The decision rule should be visible in the dashboard notes, not buried in a procedure. Leaders need to know what the indicator authorizes them to do before the meeting starts.

Step 8: Retire, redesign or keep each indicator

The eighth step is the audit decision. Each indicator should receive one of three outcomes: keep, redesign or retire. Keep means the definition is clear, the data is reliable, the owner is named, the field evidence supports the signal, and the decision rule exists. Redesign means the topic matters but the current metric does not measure it well. Retire means the indicator creates noise or distortion without useful decisions.

Retiring a metric can feel politically difficult, especially when a dashboard has displayed it for years. Yet a bloated dashboard teaches leaders to skim. A smaller set of trustworthy indicators is more valuable than a large set that rewards activity, hides weak controls, and makes every meeting feel informed while decisions stay unchanged.

In Muito Além do Zero, glossed in English as Far Beyond Zero, Andreza Araujo criticizes the way lagging indicators look in the rearview mirror and can protect the number rather than life. A weak leading indicator can create the same problem in a more modern form: it looks proactive while still protecting the report.

Leading indicator audit scorecard

The scorecard below keeps the audit practical. Use it for each indicator and avoid averaging the result too early, because one critical failure can make the indicator unsafe for executive decisions even when the total score looks acceptable.

30 days is enough to audit the first 8 to 12 indicators when the team limits the scope, samples raw records, and tests decision use rather than trying to rebuild the whole dashboard. The audit should end with a short leadership report that lists indicators to keep, redesign and retire.

Conclusion

A leading indicator quality audit protects leaders from the most dangerous metric illusion: believing that proactive data is useful simply because it arrives before an injury. Timing alone does not make a metric predictive. The indicator has to connect to exposure, survive evidence checks, and trigger decisions before the organization can call it prevention.

If your dashboard is busy but leaders still discover risk late, start with the 30-day audit above. Review 8 to 12 indicators, test their definitions, sample the evidence, walk the field, and ask what decision changed because the number moved. For organizations that want stronger safety governance, Andreza Araujo can help redesign the metric system so it supports culture, control assurance and executive decisions. Talk to Andreza Araujo about turning safety data into earlier action.