How to Test a Safety Indicator in 6 Checks Before It Enters the Monthly Dashboard

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

A new safety indicator often arrives with a convincing chart, a clean trend line, and a request to add it to the monthly leadership pack. The harder question comes before publication. What decision will change when the number moves?

If nobody can answer that question, the metric may create reporting activity without improving control reliability. This guide gives EHS and operations leaders six checks for testing an indicator before it becomes part of the monthly dashboard.

What you need before starting

Bring the proposed indicator definition, its data source, the reporting period, the intended audience, and the decision that the sponsor believes it will support. Include the current dashboard so that the review can identify duplication rather than adding another number to an already crowded page.

You also need one operational example. A metric that cannot be connected to a task, exposure, control, or management decision is not ready for approval. The review should be short, but it should include the person who owns the work represented by the data.

Step 1: Write the indicator in one precise sentence

State exactly what the indicator measures, what it excludes, the unit of measure, and the period covered. “Safety engagement” is too broad. “Percentage of scheduled critical-control verifications completed on the planned date, with evidence accepted by the control owner” can be tested.

Ask the data owner to define every important word, including completed, scheduled, accepted, and control owner. Verify the sentence against one real record and one rejected record. If two reviewers would classify the same event differently, revise the definition before collecting a trend.

The common error is allowing an attractive label to stand in for a measurable construct. A short name is useful in a chart, but it cannot replace the operational definition behind the chart.

Step 2: Confirm the denominator and collection boundary

Identify the population against which the number is calculated. A completion rate based only on tasks that supervisors remembered to schedule can look healthy while unscheduled work remains invisible. A reporting rate based on headcount may also distort the picture when exposure changes with production hours, contractor activity, or maintenance workload.

Write down the boundary in plain language. State which sites, shifts, contractors, work orders, or control checks are included, and identify the records that remain outside the measure. Compare the boundary with the operational plan for one reporting period.

Verification is complete when the denominator can be reconciled to a source record without manual interpretation. The common error is treating an easily available denominator as a valid denominator simply because the spreadsheet can calculate it.

Step 3: Link the number to a credible risk decision

Describe the decision that should change when the indicator moves. The decision may involve stopping a task, adding resources, changing a maintenance priority, escalating an overdue control, or reviewing whether a local procedure still matches the work.

Use a simple decision statement. “If the indicator falls below the agreed threshold, the area manager reviews the missed controls within five working days and decides whether work must be contained” is stronger than “leaders monitor performance.” The first statement names an action and an owner.

Test the statement with the person who has authority to act. If the owner cannot name a decision, the indicator may be descriptive rather than useful for management. That is not automatically a problem, but it should not occupy decision space on an executive dashboard.

Step 4: Test whether the data can be trusted at the required cadence

Review how the data is created, transferred, checked, and corrected. Ask whether the source is a permit system, inspection record, work-order platform, incident log, survey, or manual entry. The collection method determines what the number can reasonably claim.

Sample records from different shifts and locations, including a record that looks unusually good and one that looks unusually poor. Compare the dashboard value with the source record and document any transformation between them. If a monthly indicator depends on late corrections, show when the number becomes stable enough for a leadership decision.

Verification is complete when another reviewer can reproduce the result from the documented source and calculation. The common error is calling a number objective because it came from a system, even though users can change the input without a traceable review.

Step 5: Check the field meaning behind the trend

Take the proposed indicator to the work area or to the supervisor who manages the relevant exposure. Ask what a high result looks like in practice, what a low result looks like, and which condition could make the number misleading.

For example, a high completion rate may mean that verification is happening, or it may mean that the review is being signed before the control is tested. The chart cannot resolve that ambiguity by itself. A field check must look at the control, the evidence, and the work condition that the indicator is supposed to represent.

Record one observable confirmation and one limitation. The common error is accepting a correlation between the number and a good outcome as proof that the control is effective, even when the control has not been observed under exposure.

Step 6: Assign ownership, action, and review date

Give the indicator a business owner, a data owner, and a review owner when those roles are different. The business owner decides what changes in the operation. The data owner protects the definition and calculation. The review owner makes sure that the monthly discussion produces a documented decision.

Set a review date for the indicator itself, not only for the actions it produces. Revisit the measure when the process, exposure, system, organization, or legal requirement changes. ISO 45001, published in 2018, places emphasis on monitoring and evaluation, which means the organization should be able to explain whether its measurement remains fit for purpose.

Verification is complete when the dashboard record shows the owner, data source, decision rule, limitation, and next review date. The common error is assigning accountability to a department without naming the person who can resolve an ambiguity or stop a misleading measure.

What to do when the indicator fails a check

Do not force a weak indicator into the dashboard because the sponsor has already announced it. Mark the measure as under development, record the failed check, and identify the evidence needed to resolve it. A rejected indicator can still be useful if the review explains why it was rejected and what decision the organization needs to measure instead.

If the indicator is already in use, keep its historical values separate from the corrected series when the definition changes. Explain the break in continuity so that leaders do not read a calculation change as an operational improvement. The same discipline applies when a data source is replaced or the denominator is expanded.

For a broader method of testing whether prevention changed risk, compare this review with the safety ROI guide. When the issue is the quality of evidence reaching the monthly meeting, use the safety evidence review guide as a companion.

Approve the indicator only after the six checks have a visible result. The approval record should state the final definition, source, denominator, decision rule, limitation, owners, and review date. It should also identify the dashboard section where the measure belongs and the question it is expected to answer.

Keep the display proportional to the decision. A number that requires five explanatory footnotes may need a different design, a companion measure, or a place in the evidence appendix rather than the headline view. A dashboard becomes more useful when every visible number earns attention through a decision it supports.

Before the first monthly review, ask the operational owner to predict what action would follow a rising result and a falling result. If both answers are “we will discuss it,” the indicator is not yet connected to management action.

Final checklist for dashboard owners

Use this checklist before publishing the indicator in a monthly leadership pack.

• The definition names the measure, exclusions, unit, and reporting period.
• The denominator matches the exposure and operational boundary.
• A named decision owner can explain what changes when the number moves.
• The source and calculation can be reproduced by another reviewer.
• A field check confirms what the trend means and records one limitation.
• Business, data, and review ownership are visible.
• The indicator has a review date for changes in process, exposure, or data quality.
• The dashboard placement reflects the decision, not the sponsor's preference for visibility.

A safety indicator is ready for the monthly dashboard when it survives these checks and still helps a decision owner see what must change. If it only makes reporting look more complete, keep it out of the headline view and improve the measure first. For more guidance on executive dashboard design, read the monthly executive safety dashboard guide.