Underreporting at Work: 5 Distortions That Make Safety Metrics Look Healthy

F1 diagnostic for safety executives, EHS leaders, and operations managers

A plant can report fewer injuries while learning less about its exposure. The dashboard turns green, the monthly review becomes easier, and leaders conclude that the safety system is improving, yet the change may reflect a reporting pathway that has become harder to use rather than work that has become safer.

Underreporting is often discussed as if a worker simply decided to hide an event. That explanation is incomplete because the decision is shaped by treatment access, supervisor reactions, definitions, production pressure, and the consequences attached to the report. A reporting system is credible only when the organization makes bad news usable.

Across Andreza Araujo's work on safety culture and leadership, measurement is treated as an operating choice rather than a neutral scoreboard. The central test is therefore not whether the number is low. It is whether the number still represents what people can see, report, investigate, and change.

Why a healthy-looking metric can be weak evidence

A safety metric becomes weak evidence when the organization rewards the result more clearly than it rewards accurate reporting and effective control.

The distinction matters because outcome metrics are produced by a chain of decisions. Someone notices an exposure, decides whether it is reportable, chooses where to seek treatment, tells a supervisor, enters a record, classifies the case, and reviews the control that should prevent recurrence. A gap at any point can reduce the reported count without reducing the hazard.

James Reason's work on latent failures gives leaders a useful way to examine that chain. The missing report may be the visible symptom of a deeper arrangement involving unclear definitions, competing priorities, inaccessible care, or a manager who receives praise for a clean record and scrutiny whenever the record becomes less clean.

That is why underreporting should be examined as a measurement-system risk. The question is not whether every low number is false. The question is whether the organization has enough independent evidence to know what the number means.

Distortion 1: the reporting definition is narrower than the exposure

Every metric begins with a definition, and definitions shape what becomes visible. If the organization only emphasizes recordable injuries, people may stop discussing first-aid cases, discomfort, near misses, equipment defects, or repeated high-energy exposures that have not yet produced harm.

The narrow definition can look disciplined because it creates a clean series. It also creates a blind spot when a precursor is treated as irrelevant until the outcome crosses a compliance threshold. A worker who reports recurring hand pain, a dropped object that misses a person, or a failed isolation check may receive the message that the event is not important enough to enter the system.

Leaders should compare the formal definition with the questions used at the point of work. Can a worker report a condition before injury? Can a supervisor record a control failure without waiting for an outcome? Can the EHS team connect those entries to a critical-risk review? If the answer is no, the metric is measuring the boundary of the form, not the boundary of exposure.

OSHA recordkeeping requirements remain essential for covered employers, but compliance classification and operational learning answer different questions. A case can be non-recordable and still reveal a barrier that deserves immediate attention.

Distortion 2: treatment pathways quietly filter the data

A worker cannot report what the system makes difficult to discuss or treat. When the first response is unclear, distant, slow, or associated with an automatic investigation, people may choose informal care, continue working, or wait to see whether the condition improves.

This distortion is easy to miss because the organization may have a clinic, a hotline, and a procedure. Availability on paper does not prove that the pathway is trusted during a busy shift. A person who expects to be questioned in front of peers, removed from a preferred assignment, or blamed for a delay may use the least visible route.

The practical review should follow the event from first contact to classification. Ask who can provide immediate care, who records the concern, who decides whether work can continue, and what the worker is told about confidentiality and follow-up. A gap between those roles creates a filter before the metric is calculated.

The organization should also compare clinic visits, first-aid supplies, restricted-duty conversations, absence records, and supervisor notes with reported cases. Those sources do not prove underreporting by themselves, but a persistent mismatch is a reason to examine the pathway rather than celebrate the dashboard.

Distortion 3: supervisor incentives make disclosure expensive

Supervisors often carry the closest connection to the reporting decision. They allocate work, respond to the first disclosure, and explain what the event means for the shift. If performance conversations treat every report as evidence of poor supervision, the supervisor learns that accurate information carries a personal cost.

The resulting behavior may not look like a direct instruction to hide incidents. It can appear as delay, informal treatment, a request to wait, or a suggestion that the event was too minor to mention. Each response shifts the decision toward silence while allowing the organization to say that reporting remains encouraged.

Leaders should inspect what happens after a report, not only whether the form exists. A supervisor who reports an event should receive support to control the exposure, investigate proportionately, and close the action. If the first reaction is a performance penalty without a control review, the system teaches supervisors to protect the metric.

Andreza Araujo's book Safety Culture: From Theory to Practice is relevant because culture becomes observable in the consequences attached to information. The stated value is less important than what a supervisor experiences after bringing an uncomfortable fact into the room.

Distortion 4: case classification removes inconvenient patterns

Classification is necessary, but it can become a second reporting barrier when the conversation starts with the desired metric outcome. A review team may spend more energy proving that a case does not belong in the record than understanding why the exposure occurred.

The problem is not that every case should receive the same escalation. Proportionality matters. The problem appears when classification ends the inquiry, especially when a recurring condition is repeatedly treated as isolated, minor, or unrelated to work.

A stronger process separates at least three questions. What must be recorded or reported under the applicable rule? What credible consequence could have occurred? Which control was missing, degraded, bypassed, misunderstood, or unverified? Those questions can produce different answers, and that difference is useful evidence rather than administrative failure.

When a case is classified out of the main metric, retain the reason, the evidence, the potential severity, and the control implication. A green number without that trail is not a clean result. It is an unreviewable result.

Distortion 5: executives read absence as proof of control

Senior leaders rarely see the event that was not reported. They see a chart, a trend line, and a short explanation. If the chart improves, the organization can begin to treat the absence of bad news as proof that controls are working.

That interpretation creates a feedback loop. Confidence reduces challenge, reduced challenge leaves weak controls untested, and an untested control can remain invisible until conditions change. The metric then appears stable precisely because the organization has stopped asking what it cannot see.

Executives should require a second view of performance. Alongside injuries, review unresolved high-potential events, critical-control verification, reporting-route access, action aging, and examples in which a report changed the work. These measures are not a universal dashboard template. They are prompts that help leaders test whether the outcome number still has enough context.

The most important question may be uncomfortable. What information would make this dashboard less reassuring? If nobody can answer, the review is probably testing presentation quality instead of measurement quality.

Declared reporting culture versus operating reporting culture

A policy can say that every concern is welcome while the operating system teaches people to stay quiet. Comparing the declared arrangement with the lived arrangement makes the gap easier to investigate.

How to test the metric before trusting it

The review does not require a second dashboard with dozens of indicators. It requires a short evidence challenge that connects the number to the work.

Start with one high-consequence task or one reporting pathway. Sample recent first-aid contacts, hazard reports, near misses, restricted-duty decisions, and corrective actions. Ask the people involved where the event entered the system, what information was lost, and which decision followed.

Then test the response to an uncomfortable report. Present a hypothetical case in which a worker identifies a control weakness shortly before a production deadline. Ask the supervisor, EHS leader, and plant manager what they would do, who has authority to pause the task, and how the report would affect their performance discussion. The answers reveal the real incentive structure more clearly than a policy statement.

Finally, publish the review result as a control decision. If the pathway is weak, assign an owner, remove the access barrier, clarify classification, or change the supervisor response. If the pathway is strong, preserve the evidence that supports that conclusion and repeat the test when staffing, workload, or leadership changes.

What a credible executive conclusion sounds like

A credible safety conclusion says what the metric shows, what it cannot show, and which independent evidence supports the next decision.

Leaders do not need to reject every outcome measure. They need to stop asking one number to prove more than it can prove. A lower injury count may reflect better control, less exposure, random variation, improved recovery, incomplete reporting, or several forces at once.

The responsible conclusion therefore includes uncertainty. It names the reporting routes tested, the evidence compared, the critical controls verified, and the action required if the interpretation changes. That discipline protects the organization from both complacency and performative alarm.

Safety metrics become useful when they help leaders make a better decision about exposure. They become dangerous when leaders make the metric itself the objective.

For a practical next step, choose one green metric this month and ask what evidence would falsify the reassuring interpretation. Review the answer with operations, EHS, supervision, and workers who use the reporting route. If the number survives that challenge, confidence is more justified. If it does not, the organization has found a measurement risk before it becomes a safety event.

Explore more evidence-led safety leadership and risk analysis on the Headline Podcast blog, where technical decisions are connected to the conditions in which people actually work.