How to Audit a Leading Safety Indicator for Gaming Before the Board Review

F2 how-to guide for EHS managers, operations leaders, and board dashboard owners

A leading safety indicator can improve every month while the worksite becomes harder to understand. That happens when teams learn how to produce the number without improving the exposure, control, or decision that the number was supposed to illuminate.

The audit below is designed for the period before a board or executive review. It does not assume bad faith. It tests whether the measure still represents the work, because a metric can drift through incentives, definitions, sampling choices, or simple routine.

What you need before starting

Bring the indicator definition, denominator, data owner, collection instructions, recent trend, and the decision forum in which the result is reviewed. Add a small sample of source records and at least one field check that tests what the number claims to represent.

Use the same discipline applied in a safety indicator validation review, but add one question that is often missing: what behavior does this measure make rational when people are under pressure?

Step 1: State the decision the indicator must change

Write the decision in one sentence before looking at the trend. For example, the review may decide whether to redesign a critical control, change supervision, fund an engineering improvement, or investigate a new exposure. If the only stated purpose is to show commitment, the measure is not ready for board use.

Name the decision owner and the threshold that triggers action. A measure that has no owner, no response, and no review date becomes a communications asset rather than a management control. Compare the result with the questions used in an executive safety dashboard that changes decisions.

Step 2: Separate activity from evidence of control

Divide the indicator into the activity being counted and the evidence that the activity changed risk. A completed observation, training session, inspection, or meeting is activity. Evidence may include a corrected control defect, a verified barrier, a changed work method, or a decision that removed an exposure.

This distinction does not make activity irrelevant. It prevents activity from being treated as proof of effectiveness. ISO 45001 requires organizations to monitor and evaluate relevant processes, while the evaluation still depends on what the process reveals and what the organization does with that information.

Step 3: Test the denominator and classification rules

Read the definition as if you were trying to produce a favorable result without falsifying a record. Ask which events are excluded, who decides whether an observation is valid, whether the denominator changes with staffing or production, and whether a late entry is counted in the period when the work occurred.

Then compare the rules with the recordkeeping logic used for injury and illness data. OSHA recordkeeping definitions are not a universal template for leading indicators, but they show why classification choices must be explicit, consistent, and reviewable. A trend that improves after the classification rule changes is not comparable until the break is disclosed.

Step 4: Sample the records behind the score

Select records from a favorable month, an average month, and a month in which the number moved unexpectedly. Check whether entries are specific enough to identify the task, exposure, control, owner, and follow-up. A high count made of vague statements is not strong evidence of prevention.

Ask the frontline owner to explain how the record was created, what happened after submission, and whether a concern could have been recorded without creating friction. If the process rewards easy entries and makes difficult findings costly, the score is measuring participation under pressure rather than control quality.

Step 5: Compare the number with field conditions

Take a sample of the exposure or control that the indicator claims to monitor and verify it at the point of work. Look for agreement between the dashboard, the work method, the condition of the equipment, and the worker's description of what happens when the plan changes.

Disagreement is not automatically proof that the data is false. It is evidence that the indicator has a boundary. Record the mismatch, identify whether the problem sits in collection or control execution, and avoid averaging the exception away. James Reason's distinction between active failures and latent conditions is useful here because a clean report can coexist with organizational conditions that make weak controls likely.

Step 6: Rehearse the board question before the review

Ask a leader who did not design the measure to challenge it. The reviewer should be able to answer what the indicator shows, what it does not show, which behavior it may encourage, and what decision follows from a favorable or unfavorable result.

Put limitations beside the trend, not in a footnote that nobody reads. If the measure rewards activity instead of control, say so and pair it with a more direct signal. The board should see the relationship between exposure, control effectiveness, and decision latency, rather than receiving a single score that implies more certainty than the evidence supports.

Final checklist before the board review

Record the indicator's decision purpose, definition, denominator, data owner, sampling method, field-verification result, known gaming risk, and next review date. Keep the original version of the definition whenever a rule changes so that trend breaks remain visible. If the audit identifies a weak measure, keep it with a clear limitation, redesign it so collection tests control quality, or retire it in favor of evidence that can change an operational decision.

• The indicator has a named decision owner and a defined response to movement.
• Activity is separated from evidence that exposure or control reliability changed.
• Definitions, exclusions, denominators, and classification rules are stable or disclosed when changed.
• Source records have been sampled across favorable, ordinary, and unexpected periods.
• At least one field verification tests whether the trend matches the work.
• Known limitations and possible gaming behaviors appear beside the result.

Conclusion: make the metric earn its authority

A leading indicator earns authority when improving the number requires better evidence, stronger controls, or a clearer decision. It loses authority when people can improve the report while leaving the exposure untouched.

Andreza Araujo's work on safety culture keeps this distinction practical. Across her executive experience and in Safety Culture: From Theory to Practice, the question is not whether an organization can produce a safety signal. The question is whether the signal changes how work is led. For more guidance, read about leading indicators that reward activity instead of control and visit Andreza Araujo.