Workplace Bullying: 8 Blind Spots in an Investigation

A workplace bullying investigation becomes credible when it examines conduct, power, work design, evidence, and retaliation risk together. The central mistake is treating the report as a dispute between two personalities, because the organization may have created the conditions that let the conduct continue.

ILO Convention No. 190 was adopted in 2019 and entered into force in 2021, recognizing that violence and harassment can be a single occurrence or repeated conduct that causes, or is likely to cause, physical, psychological, sexual, or economic harm. This article identifies 8 investigation blind spots that keep leaders focused on labels instead of controllable facts.

Why a bullying report is also a work-system signal

Workplace bullying is often discussed as if the organization must first decide whether the label is legally correct. That sequence is too narrow. A report can reveal humiliation, intimidation, exclusion, threats, unreasonable control, retaliation, or chronic role conflict before the facts fit a specific legal category.

The investigation therefore has two connected questions. What happened between the people involved, and which management decisions allowed the exposure to persist? The second question is where many reviews become decorative, because leaders close the case after reaching a view about intent while leaving the schedule, reporting line, incentive, or supervisor behavior unchanged.

Andreza Araújo makes this distinction practical in Safety Culture Diagnosis: Learn how to do your own, where diagnosis is useful only when it leads to a clearer decision about the conditions that need to change. A bullying investigation deserves the same discipline. The record should support protection, accountability, and prevention rather than a faster administrative ending.

1. Treating the report as a personality conflict

The first blind spot appears when a complaint is translated into “two people do not get along.” That wording removes the conduct from the record and makes the reporting person carry the burden of proving a relationship problem instead of describing a work exposure.

Investigators should define the behavior in observable terms. “The supervisor is toxic” is weak evidence. “The supervisor removed the employee from 3 shift meetings after the employee questioned a production decision” is a fact pattern that can be checked against calendars, messages, and witness accounts.

Across 25+ years leading EHS work, Andreza Araújo has consistently connected culture to visible decisions rather than slogans. Apply that lens here by asking who controlled access, information, assignments, performance ratings, and escalation routes. The power relationship may be more important than the personalities.

When the report is rewritten as conduct and decision history, the investigation can test recurrence, intent, impact, and organizational responsibility without assuming that every difficult interaction is bullying.

2. Using the legal threshold as the starting gate

The second blind spot is waiting for a report to meet a legal definition before management responds. The EEOC describes unlawful harassment through conduct that becomes a condition of continued employment or is severe or pervasive enough to create an intimidating, hostile, or abusive environment. Those are important legal tests, but they are not a complete risk-management screen.

OSHA describes workplace violence broadly as acts or threats of physical violence, harassment, intimidation, or other threatening behavior, while noting that the United States has no specific OSHA workplace-violence standard and that the General Duty Clause appears in Section 5(a)(1) of the Occupational Safety and Health Act of 1970. The practical lesson is not that every rude interaction is an OSHA case. It is that leaders should not use the absence of a single label as permission to ignore a recognized hazard.

Separate the legal assessment from the management response. A report may require legal review, interim protection, a conduct correction, a work-design review, or all 4 in sequence. The decision should document which question each action answers.

This approach prevents two errors at once. It avoids casual legal conclusions, and it stops the organization from treating “not proven as unlawful harassment” as “nothing requires attention.”

3. Collecting opinions instead of a time-bounded record

A credible investigation needs a defined time window, because a vague question produces vague memories. Start with the first reported event, the most recent event, and any management decision that changed the person’s exposure, access, status, or opportunity.

Use a record that separates 4 evidence types: direct observations, contemporaneous documents, repeated patterns, and impact indicators. Messages and schedules can establish timing. Witness accounts can confirm language or exclusion. Absence, turnover, transfer, or performance changes may signal impact, but they do not prove causation by themselves.

The ILO definition in Convention No. 190 is useful because it includes conduct that is likely to result in harm, not only harm that has already been medically documented. That supports early action while preserving the distinction between an exposure signal and a clinical diagnosis.

For the investigator, the practical test is simple. Can another qualified reviewer understand what was alleged, which evidence was checked, what remains uncertain, and why each finding was reached? If not, the file is an opinion bundle rather than an investigation record.

4. Interviewing only the two central people

Investigating only the complainant and the subject of the complaint creates a false binary. Workplace conduct travels through meetings, work allocation, handovers, chat groups, performance systems, and informal access to decision-makers, so the relevant evidence may sit with people who were not the visible participants.

Build an interview map before scheduling conversations. Include people who observed the interaction, received similar treatment, approved the work decision, handled an earlier concern, or can explain the normal process. A witness does not need to agree with the complainant to provide useful evidence.

Keep the questions behavior-specific. Ask what was said or done, when it happened, who was present, what decision followed, and whether the same pattern appeared elsewhere. Avoid asking witnesses to decide whether the conduct “counts” as bullying, because that invites labels before facts.

A wider evidence map also reduces retaliation risk. When the organization has documented the work context rather than relying on one person’s credibility, it is harder for a powerful actor to recast the report as an isolated personal grievance.

5. Ignoring power, dependency, and retaliation

Power is not limited to the formal reporting line. A project lead, scheduler, subject-matter expert, senior peer, contractor representative, or customer-facing manager may control access to shifts, overtime, information, assignments, or future opportunities.

ILO Convention No. 190 applies across public and private sectors and formal and informal economies. That breadth matters because a narrow employee-only lens can miss contractor interfaces, temporary work, remote channels, and third-party conduct that still affects the work environment.

Ask what the reporting person could lose by speaking. The answer may include a preferred shift, a promotion path, a contract extension, a desirable assignment, or the informal support needed to perform safely. Interim safeguards should reduce that dependency without automatically moving the reporting person away from the work or treating them as the problem.

Retaliation monitoring needs a date, an owner, and observable checks. A promise to “keep an eye on things” is not a control. Review assignment changes, meeting access, feedback, performance actions, and communication patterns at 7, 30, and 60 days when the risk warrants those checkpoints.

6. Separating conduct from work design

The sixth blind spot is investigating the behavior while ignoring the work that amplifies it. Chronic understaffing, conflicting priorities, impossible deadlines, unclear authority, and unstable schedules can create repeated pressure in which humiliation or intimidation becomes normalized as a management method.

The World Health Organization reported in 2023 that an estimated 16% of working-age adults were living with a mental disorder, while depression and anxiety contribute to an estimated 12 billion lost workdays each year at a global productivity cost of about US$1 trillion. Those figures do not prove that bullying caused a person’s condition. They show why organizations should treat mental health and work design as material management concerns, not private weakness.

Use the investigation to ask which operating condition made the conduct easier to repeat. Did a target force supervisors to compete for output? Did the roster leave one person isolated? Did a manager lack an escalation path? Did performance language reward public pressure? A behavior correction without a work-design correction may only move the exposure to the next shift.

ISO 45003:2021 provides a useful structure for identifying psychosocial hazards within an occupational health and safety system. The standard does not replace fact-finding, but it helps leaders connect interpersonal conduct with demands, control, support, relationships, and organizational change.

7. Treating confidentiality as absolute secrecy

People often hear “confidential investigation” and understand that no one else will know. That promise is rarely realistic. A fair process may require information to be shared with decision-makers, investigators, legal advisers, occupational health professionals, or witnesses who need enough context to answer questions.

The safer commitment is controlled information sharing. Explain who will receive information, why they need it, how records will be protected, and what the organization can do if retaliation occurs. The investigator should disclose only what the next decision requires, rather than circulating the complete complaint.

This boundary protects both sides. The reporting person receives a process that is serious without being exposed to unnecessary gossip, and the subject receives a fair opportunity to respond to specific allegations rather than a reputation built from partial information.

Document the privacy decision as carefully as the factual decision. If the case file cannot show why information was shared, withheld, or redacted, trust will be damaged even when the final finding is reasonable.

8. Closing the file when the action is assigned

Assignment is not control. A corrective action that says “coach the manager” may look complete in a tracker while the same meeting behavior, work pressure, and power imbalance remain unchanged.

Each action should name the exposure, the owner, the due date, the evidence of completion, and the verification method. “Conduct training completed” is one record. “Meeting access restored, work allocation reviewed, retaliation checks completed at 30 and 60 days, and new complaints assessed” is a control-verification plan.

The organization should also decide how it will communicate the outcome without disclosing protected details. People do not need every fact in the case, but they do need to know that the concern was reviewed, that safeguards exist, and that reporting a recurrence will trigger a response.

Andreza Araújo’s safety culture work emphasizes that care becomes credible when it appears in decisions. The same principle applies here. A closed file is not evidence of a safe workplace; a verified change in conduct and conditions is stronger evidence.

Declared investigation versus protective investigation

The comparison is not between a soft process and a strict one. It is between a process that protects the organization’s paperwork and one that protects people, evidence, and the conditions required for safe work.

What leaders should decide next

A workplace bullying investigation should produce one clear management sentence: what conduct or condition requires control, who owns the response, and how the organization will verify that the exposure has changed. The 8 blind spots above matter because each one can make a harmful pattern look smaller, more private, or less actionable than it is.

For practitioners who need to connect psychosocial risk, leadership behavior, and operational reality, Safety Culture: From Theory to Practice and Safety Culture Diagnosis: Learn how to do your own provide Andreza Araújo’s broader framework. Explore the Andreza Araújo book collection when your next step is to turn diagnosis into visible care. Safety is about coming home.