How Cam Stevens Thinks About When a Safety Metric Earns Trust

Episode 15 of the Headline Podcast, published on March 18, 2026, features Cam Stevens discussing technology, prediction, and the safety problems leaders are trying to solve. His central argument is useful for measurement because a metric should earn trust by improving a real decision, not by making the organization feel better informed.

A senior team can approve a dashboard in an afternoon and still remain unable to answer the question that matters most. What will we do differently when this number changes? Cam's problem-first reasoning turns that question into the starting point for measurement design.

What Episode 15 changes about safety measurement

Many measurement programs begin with the data that is easiest to collect. The organization counts observations, inspections, completed actions, training attendance, or alerts, then asks leaders to infer whether the system is becoming safer. That order is convenient, although it reverses the logic of control.

Cam Stevens said on the podcast, "If we're very clear on the problem to solve, then a technology catalog is excellent. If we're not clear on the problem to solve, a technology catalog is very dangerous." The same distinction applies to metrics. A measure can be technically accurate and still be irrelevant to the decision it is expected to support.

The practical implication is that a metric review should begin with a decision statement. The team should name the exposure, the decision owner, the action that may change, and the evidence that would show whether the action worked.

Why the problem must come before the metric

Suppose a board wants a single safety score for a group of sites. The request sounds reasonable until the team asks what the score is meant to decide. Is it intended to allocate capital, identify a site that needs support, verify a critical control, compare leadership follow-through, or reassure investors?

Those decisions require different evidence. A rate that helps compare broad performance may be too slow for a control-health decision. An inspection result that helps a supervisor act during a shift may be too narrow for capital allocation. A metric becomes misleading when one number is asked to answer all of those questions at once.

Leaders should therefore write the problem in one sentence before approving a measure. The sentence should describe a condition that can change, not a desire to improve visibility. "We need better safety data" is not a problem statement. "We need to know which overdue critical-control checks require executive intervention this week" is specific enough to guide design.

What a trustworthy safety metric actually measures

A trustworthy metric does not need to predict every incident. It needs to represent a decision-relevant condition with enough clarity that a named person can act and verify the result. Four questions expose whether the measure has that quality.

The last question matters because a metric is still an interpretation. A completed check can mean that a form was submitted, not that a barrier worked. A low number of reported concerns can mean that the workplace is quiet, or it can mean that people have learned reporting is costly. Verification keeps the measure connected to what happened outside the screen.

This is why the four tests for safety leading indicators should be applied before a measure is promoted to the executive dashboard.

How dashboards create false confidence

Dashboards create false confidence when visual certainty outruns operational certainty. Green status, complete data, and a favorable trend can coexist with a control that is unavailable, misunderstood, or bypassed under production pressure.

The distortion often begins with aggregation. Different sites, shifts, hazards, and response times are combined because the executive audience wants a concise view. The result may be easy to read, but the aggregation can hide the one condition that requires intervention.

Another distortion appears when the measure rewards the act of recording rather than the quality of the decision. Teams learn to close actions, complete observations, or resolve alerts quickly because the system recognizes closure. The underlying exposure may remain unchanged while the score improves.

The verification debt created by paper controls is a useful warning for metric owners. If the organization cannot show how a result was checked in normal work, it should not present the number as proof that the control is healthy.

Where human judgment stays in the measurement system

Cam Stevens's technology discussion does not reject data. It asks leaders to protect the human judgment that data cannot carry. A metric can indicate that something changed, but it cannot always explain why the change matters or which response is proportionate.

Consider a rise in overdue inspections. The number may indicate a staffing problem, a scheduling failure, a permit restriction, or a deliberate pause because the inspection method is not safe under current conditions. Treating every increase as a compliance failure would produce the wrong intervention.

Human judgment is also needed when evidence conflicts. A dashboard may show that a control is available while a worker reports that the control cannot be used during a specific task. The disagreement is not noise to remove. It is a decision point that requires someone with authority to test the condition and decide what protection applies while the gap remains open.

In Safety Culture: From Theory to Practice, Andreza Araujo's work keeps culture connected to the decisions people make under pressure. A measurement system should follow the same principle. Its purpose is not to eliminate judgment, but to make judgment better informed and more accountable.

A 30-day review for one executive metric

Choose one metric that senior leaders currently trust and review it without defending its history. The objective is to learn whether the measure changes the condition it is supposed to represent.

1. Days 1 to 3: write the problem statement, the decision, the owner, and the action that should change when the metric moves.
2. Days 4 to 10: observe the related work in at least two normal operating conditions and record what the metric cannot see.
3. Days 11 to 17: compare the reported result with field evidence, worker knowledge, records, and the actual status of the control.
4. Days 18 to 24: ask the decision owner what action the metric triggered, what action it failed to trigger, and which incentives may distort reporting.
5. Days 25 to 30: decide whether to keep, redesign, or retire the metric, then document the evidence behind that decision.

The review should not become another reporting exercise. Its output is a decision record that says what the metric represents, where it is reliable, where it is weak, and what leaders must do when the signal is incomplete.

The four states of a safety signal offer a practical way to distinguish an observation from an operating decision, which prevents a dashboard from becoming the final step in the control process.

Recommendation for C-suite and EHS leaders

Executives should stop asking whether a safety metric is leading or lagging before asking whether it changes a decision. The label matters less than the connection between the signal, the exposure, the owner, and the response.

Cam Stevens's Episode 15 argument provides a disciplined order. Define the problem, identify the signal, test the signal against work, protect human judgment, and scale only when the metric improves action without creating a new burden for the people who must produce or interpret it.

That approach also changes the leadership conversation. Instead of asking why a site is green, the executive asks which condition the green status represents, who verified it, and what would make the status unreliable. Those questions create a more honest dashboard because they keep uncertainty visible.

Listen to the full conversation with Cam Stevens on the Headline Podcast. The episode is a useful starting point for leaders deciding whether a new metric will improve risk decisions or only improve the appearance of control.

Conclusion: a safety metric earns trust when leaders can name the problem it represents, connect it to a decision, verify it against work, and act when the signal exposes a condition that the dashboard cannot resolve alone.