How to Measure Safety Culture ROI in a 90-Day Pilot

A safety-culture program can consume budget for months while its only visible output is a higher training attendance rate. That is not evidence of return. It shows that people were present in a room, not that critical decisions became safer when work conditions changed.

Safety-culture ROI is the measurable relationship between an investment in culture and a change in exposure, control reliability, reporting quality, or decision speed. A useful 90-day pilot therefore links field evidence to management action, while keeping injury counts in their proper place as lagging outcomes rather than the only proof of value.

The method below is designed for an EHS manager who needs to explain value to operations and finance without turning safety into a promise that every incident can be prevented. Andreza Araújo's Safety Culture Diagnosis: Learn how to do your own treats indicators as evidence for diagnosis, while Far Beyond Zero, her Portuguese title Muito Além do Zero, warns that a clean lagging number can conceal weak conditions or underreporting.

What you need before starting

Choose one operational problem with a visible decision boundary. Examples include overdue critical-control actions, weak near-miss follow-up, repeated permit deviations, or maintenance work that starts before hazards are verified. Do not begin with the broad promise to improve culture. Begin with a specific exposure that leaders can influence within one quarter.

You also need a named business owner, a baseline period, access to the existing records, and permission to report an adverse signal without editing it to preserve a green dashboard. The pilot becomes credible when operations owns the decision and EHS owns the measurement discipline.

Step 1: Choose the decision that should improve

Write the management decision in one sentence. It might be, “The maintenance manager will not release non-routine work until the critical controls have current evidence.” Another example is, “The plant manager will resource corrective actions that remain open because the assigned team lacks capacity.”

Verify this step by asking who can make the decision, what information they receive, and what would cause them to change course. The common error is selecting a theme such as engagement or awareness, which sounds important but does not identify a decision that can be observed.

Step 2: Define the investment boundary

List the resources that the pilot will use, including manager time, field coaching, data preparation, supervisor release time, specialist support, and any technology or contractor cost. Include internal time because a program that appears free can still displace inspections, maintenance planning, or production coordination.

Keep the boundary narrow enough to calculate. If the pilot includes every safety activity in the facility, the result will become a general budget narrative rather than a test of whether one intervention changed one class of decisions. Verify the list with finance or the operational sponsor before the pilot starts.

Step 3: Set a baseline that describes conditions

Capture the current state before the intervention begins. Use a short period that represents normal work, then record the starting values for the selected exposure, the age of open actions, the proportion of evidence that is current, the quality of near-miss follow-up, and the time between a reported signal and a management decision.

Do not invent a financial value for every observation. Some changes are best expressed as avoided exposure hours, restored control availability, or faster escalation. Verify the baseline by sampling records against field conditions, because a perfect database can still describe work that no longer exists.

Step 4: Select one outcome and three leading signals

Choose one outcome that matters to the operation, such as fewer overdue critical actions, fewer repeat exposures, or a shorter time to resolve a verified barrier failure. Then select three leading signals that explain how the outcome may change. A practical set includes the percentage of sampled controls with current evidence, the percentage of near-miss actions closed with a verified change, and the median time from signal to decision.

Each signal needs a definition, owner, collection frequency, and review threshold. A signal becomes useful when its red condition changes a decision. Verify the set by removing any measure that only reports activity, such as the number of meetings held, unless meeting quality is independently assessed.

Step 5: Connect each signal to a control question

Translate every metric into a question that a supervisor or manager can answer. Instead of asking whether the completion rate is 92 percent, ask whether the control was available at the moment the exposure existed, whether its failure was escalated, and whether the assigned action changed the work.

This step protects the pilot from false precision. A percentage can look stable while the highest-consequence tasks remain unverified. Compare the result with the decision tests in this safety-dashboard guide before accepting a comfortable number. Verify the connection by reviewing a small sample of records and asking whether the metric would have exposed a decision failure that the operation already recognizes.

Step 6: Run a weekly evidence review

Hold a short weekly review with the operational owner, the relevant supervisor, and the EHS measurement lead. Review the signal, the supporting evidence, the current exposure, the decision taken, and the next verification date. Keep the discussion close to the work so that the dashboard does not become a separate reporting ritual.

Record disagreements instead of smoothing them away. A supervisor may believe a control is adequate while the evidence shows that it is unavailable during a specific shift or contractor interface. Verify this step by checking that every red signal has an owner and a stated decision, even when the decision is to gather better evidence.

Step 7: Convert changes into a defensible value story

At the halfway point, describe what changed in operational terms. The value story might be that a high-risk job was delayed until an isolation was verified, that an action backlog was reduced by removing a resource constraint, or that a repeated exposure became visible before an injury occurred.

Separate direct savings, avoided exposure, and capability gains. Direct savings may be documented through reduced rework or contractor delay. Avoided exposure should remain an explicit risk-based interpretation, not a claimed injury that did not happen. Capability gains include faster escalation and better evidence, which can strengthen future decisions without pretending that every gain has a precise dollar value.

Step 8: Test whether the result survives a red signal

A culture measure is not credible if it only rewards green results. Deliberately examine one adverse signal and ask whether the organization challenged the number, protected the reporter, and changed the work where necessary. Andreza's grounding on indicators makes this distinction clear. Not every green result is success, and not every red result is failure when the red signal reveals a condition that leaders can finally address. The same discipline applies to zero-accident targets, which can reward silence when leaders protect the number instead of the work.

Verify the test by looking for an actual decision after the red signal. If the organization simply changes the target, closes the record without evidence, or blames the reporter, the pilot is measuring dashboard protection rather than culture improvement.

Step 9: Compare the pilot with the baseline

At day ninety, compare the same definitions, scope, and sampling logic used at the start. Show the outcome, the three leading signals, the investment boundary, the decisions that changed, and the unresolved limitations. If the data quality improved but the exposure did not, report that result without forcing a success narrative.

Verify comparability before interpreting movement. A metric can improve because the team changed the definition, excluded difficult work, or stopped recording weak signals. The pilot has value when it makes those distortions visible and gives leaders a better basis for the next decision.

Step 10: Decide what to scale, stop, or redesign

End with three decisions. Scale the parts that changed the target decision and produced reliable evidence. Stop activities that only create attendance or document volume. Redesign measures that were technically correct but too slow, too distant from the work, or too easy to manipulate.

Present the recommendation to the operational owner, not only to the safety committee. A safety-culture investment earns credibility when leaders can explain what they changed, what exposure became visible, and what they will resource next. That is a stronger return story than a list of completed workshops.

Final checklist for a 90-day safety-culture ROI pilot

• Name one operational decision that should improve.
• Define the investment boundary, including internal time.
• Record a baseline using the same definitions that will apply at day ninety.
• Choose one outcome and three leading signals.
• Connect each signal to a control question and an owner.
• Review evidence weekly and preserve disagreement.
• Separate direct savings, avoided exposure, and capability gains.
• Test the system with a red signal before declaring success.
• Compare like with like and document data limitations.
• Decide what to scale, stop, or redesign.

The purpose of a safety-culture ROI pilot is not to make human protection look like a short-term sales case. It is to show whether investment changes the decisions that keep exposure visible and controls reliable. When the measurement stays close to the work, leaders can see value without hiding uncertainty behind a green number.

Build a more practical safety culture. Explore Andreza Araújo's work on safety culture, leadership, and prevention.