Near Miss Report Form: Essential Robot Incident Guide 2026

A near miss report form is a short, structured document that captures an event in which a worker might have been hurt if the circumstances had been slightly different. Its purpose is to get the hazard investigated and fixed before it injures someone. At minimum it must record the who, what, when, and where of the event, the potential severity, and the immediate actions taken. For robot incidents it must also capture the robot system state: model, operational mode, software and firmware version, and safety-system status.

That last part is what most EHS teams get wrong. I have reviewed plenty of generic near miss forms that work fine for a slip on a wet floor and fail completely when a humanoid robot lurches during a maintenance task. This guide covers the standard form structure first, then the robot-specific fields, the 5-Why method with a worked robot example, the 29 CFR 1904 recordability rules, and the workflow that keeps the whole system honest. It is the second part of my larger robot safety documentation guide.

What Is a Near Miss and Why Must You Report It?

OSHA defines close calls, also called near misses, as incidents “in which a worker might have been hurt if the circumstances had been slightly different” (OSHA). The agency states that investigating incidents and near misses lets employers “identify and implement the corrective actions necessary to prevent future incidents.” A near miss is free information. Nobody got hurt, yet the hazard revealed itself.

The economic case is blunt. Work injuries cost the US economy $176.5 billion in 2023, and the average cost per medically consulted work injury was $43,000 (National Safety Council, 2023 data). Every near miss you investigate and close is a chance to avoid that bill.

For robot operations the stakes are documented in fatality data. A NIOSH analysis identified 41 robot-related workplace fatalities in the US between 1992 and 2017. In 78 percent of those cases the robot struck the worker, and 83 percent involved stationary robots (NIOSH; Layne, AJIM 2023). Struck-by events rarely come out of nowhere. They are usually preceded by warning signs that a working near miss program would have caught.

One caution. You may have seen the Heinrich triangle, the old claim that a fixed ratio of near misses sits beneath every serious injury. Treat it as a disputed historical heuristic, not established fact. The real argument for reporting does not need an invented ratio: each near miss is a specific hazard, observed in your facility, that you can fix before it bites.

OSHA defines close calls as incidents “in which a worker might have been hurt if the circumstances had been slightly different” and notes that investigating them lets employers “identify and implement the corrective actions necessary to prevent future incidents.”

What Fields Belong on a Near Miss Report Form?

A usable near miss report form fits on one page and takes under ten minutes to complete. If it takes longer, people stop filing reports, and your data dries up. Here are the core fields every form needs, robot cell or not, and why each one matters.

Field	Why It Matters
Date, time, exact location	Lets investigators pull the right shift records, camera footage, and robot logs.
Reporter name and role (anonymous option)	Follow-up questions need a contact. An anonymous channel keeps reports flowing where trust is low.
Plain-language event description	Facts only, no blame. What happened, in the order it happened.
People involved and witnesses	Memory fades fast. Witness lists collected on day one save the investigation.
Potential severity rating	A near miss that could have killed someone gets a different response than a stubbed toe avoided.
Immediate actions taken	Shows whether the hazard is contained right now or still live on the floor.
Contributing conditions	Lighting, housekeeping, time pressure, staffing. These feed the root cause analysis.
Photo, video, or log file reference	Evidence attached at report time is evidence that does not get overwritten later.
Supervisor notification timestamp	Proves the escalation chain worked and starts the 48-hour clock with a verifiable time.

Keep the severity rating simple: a three-level scale such as minor, serious, and potentially fatal works better in practice than a ten-point matrix nobody agrees on.

What Robot-Specific Information Must the Form Capture?

Here is where generic forms fail. When a robot is involved, the investigator’s first question is always the same: what was the machine doing at that moment? If the form does not capture the robot system state, the event often cannot be reconstructed, because logs roll over, firmware gets updated, and the cell configuration changes before anyone looks.

Add these fields to the robot section of your near miss report form:

• Robot identification. Make, model, and serial number. “The arm in cell 4” is useless twelve months later when the fleet has been reshuffled.
• Operational mode at the time of the event. Automatic, manual reduced speed, teach mode, or maintenance mode. The same motion means completely different things in different modes.
• Software and firmware version. Behavior changes between releases. An investigator who cannot pin the version cannot rule a software regression in or out.
• Safety-system status. Which safeguards were active: emergency stop circuits, light curtains, area scanners, safety-rated monitored functions. Were any in a muted or bypassed state, and was the bypass authorized?
• Task and payload. What program was running and what the robot was carrying.
• Stability state for mobile and legged platforms. For humanoids and other dynamically stable robots, record whether the robot was standing, walking, supported by a fixture, or powered down. The standards work I covered in my review of humanoid robot safety standards exists precisely because these machines can become unstable when power is removed.
• Log file capture confirmation. A checkbox confirming that controller logs and any video were exported and attached. This single checkbox has saved more investigations than any other field I know.

The NIOSH fatality data makes the case for this level of detail. With 78 percent of the 41 documented robot-related deaths involving a robot striking the worker (NIOSH), reconstructing exactly what the machine was commanded to do, and which safeguard should have stopped it, is the heart of every robot incident investigation.

How Does a 5-Why Root Cause Analysis Work?

The 5-Why method asks “why did that happen?” repeatedly, usually about five times. Each round moves you past the obvious trigger and closer to the system failure underneath. It works because the first answer is almost never the real cause. The first answer is usually a person; the real cause is usually a process.

Here is a worked robot example, anonymized from the kind of event I hear about in humanoid pilot programs.

Event: During scheduled maintenance, a bipedal robot started tipping when a technician released a joint brake. The technician jumped clear and the robot was caught by a colleague with a strap. Nobody was hurt. Potential severity: serious.

1. Why did the robot start tipping? A joint brake was released while the robot was standing unsupported, and the leg folded under the robot’s own weight.
2. Why was the robot unsupported? The bracing fixture step was skipped before brake release.
3. Why was the bracing step skipped? The technician did not know bracing was required before releasing that brake.
4. Why did the technician not know? The energy control procedure that lists the bracing step was not posted at the cell, and the technician worked from memory.
5. Why was the procedure not posted? The cell setup checklist never required posting the energy control procedure. Posting was assumed, never assigned.

Root cause: a missing requirement in the cell setup checklist, not a careless technician. The corrective actions follow directly. Add the posting requirement to the checklist, audit every existing cell against it, and brief all technicians on the change. Note how the analysis lands on the documentation system. That is typical, and it is why a robot-specific lockout/tagout procedure that addresses stored energy and stability is not optional paperwork. OSHA’s control of hazardous energy standard, 29 CFR 1910.147, requires documented energy control procedures for exactly this kind of servicing work (OSHA).

Two rules keep 5-Why honest. Stop when the answer becomes a fixable system condition, whether that takes three whys or seven. And never accept “human error” as a root cause; ask why the system allowed the error to matter.

When Is an Incident OSHA-Recordable Under 29 CFR 1904?

A near miss, by definition, is not recordable. OSHA’s recordkeeping rule, 29 CFR Part 1904, covers work-related injuries and illnesses, and recording is generally triggered when a work-related event results in death, days away from work, restricted work or job transfer, medical treatment beyond first aid, or loss of consciousness (OSHA, 29 CFR 1904). No injury, no entry on the OSHA 300 log.

So why does recordability belong in a near miss guide? Three reasons.

First, the line moves fast. The same pinch event that produced a bruise today produces a recordable laceration tomorrow. Your form should prompt the reporter to flag whether any medical evaluation occurred, because “medically consulted” is exactly where costs start: $43,000 on average per medically consulted work injury (National Safety Council, 2023 data).

Second, severe outcomes carry separate, fast reporting duties under OSHA’s severe injury reporting rules in 29 CFR 1904.39. A work-related fatality must be reported to OSHA within 8 hours, and an inpatient hospitalization, amputation, or loss of an eye within 24 hours. Your escalation flowchart needs those clocks printed on it.

Third, a near miss file is your best defense and your best teacher. If an inspector ever asks what you knew about a hazard, a documented near miss with closed corrective actions shows a functioning safety system. The same file is the raw material for the robot safety training program that keeps new hires from repeating old close calls.

When in doubt about a specific case, read the standard itself rather than a summary. Part 1904 is short and unusually readable as regulations go.

How Do You Run the 48-Hour Reporting Workflow?

The discipline I recommend is simple: verbal notification before the end of the shift, written initial report within 48 hours, no exceptions. The 48-hour window is not a regulatory deadline for near misses. It is an operational one. After two days, memories blur, log buffers overwrite, and the cell has usually been reconfigured.

The full workflow runs in six stages, shown below.

A few practical notes on running it. The initial report records facts, not conclusions; root cause comes later, in stage four. The supervisor who receives the verbal notification owns the 48-hour clock and confirms the robot logs were exported before anything is reset. And the investigation team should include someone who can read those logs, because a robot near miss investigation without controller data is guesswork.

Make the form trivially easy to reach: paper copies at the cell, a QR code on the safety board, a shared drive link. Friction kills reporting programs faster than fear does.

How Do You Track Corrective Actions to Closure?

A near miss report without a closed corrective action is a filed complaint, nothing more. OSHA’s incident investigation guidance is explicit that the point of investigating near misses is to “identify and implement the corrective actions necessary to prevent future incidents” (OSHA). Implementation is the part most programs skip.

The tool is a corrective action log with four mandatory columns:

• Action. Specific and verifiable. “Add energy control procedure posting to the cell setup checklist,” not “improve awareness.”
• Owner. One named person. Shared ownership is no ownership.
• Due date. Realistic, tracked, and escalated when missed.
• Status. Open, in progress, completed, or verified. The gap between completed and verified matters: completed means someone says the fix is done, verified means someone independently checked that it works on the floor.

Review the log in your regular safety meeting and close items only after verification. In the tipping-biped example above, closure means a supervisor walks every robot cell, finds the energy control procedure posted, and confirms the updated checklist is the one technicians actually use. Trend the log quarterly too. If the same robot model or the same operational mode keeps appearing, you have found a fleet-level issue worth raising with your integrator.

Frequently Asked Questions

What is the difference between a near miss and an incident?

An incident causes actual harm or damage; a near miss is an event in which a worker might have been hurt if the circumstances had been slightly different, which is OSHA’s own definition (OSHA). Both deserve investigation, because the underlying hazard is identical. Only the outcome differs.

Are near misses recordable under OSHA 29 CFR 1904?

No. Recording under 29 CFR Part 1904 is triggered by work-related injuries and illnesses meeting criteria such as medical treatment beyond first aid or days away from work (OSHA). A near miss involves no injury, so it is not recordable. Document it anyway; that file drives prevention.

Why do robot near miss forms need the software version?

Robot behavior can change between software and firmware releases. Without the exact version, investigators cannot determine whether a regression, a configuration change, or a hardware fault produced the event, and they cannot check whether other robots in the fleet run the same vulnerable version. It is one line on the form.

How fast should a near miss be reported?

My standard is verbal notification to the supervisor before the end of the shift and a written report within 48 hours. That window is operational, not regulatory: controller logs get overwritten, witnesses forget details, and cells get reconfigured. Severe injury outcomes have separate OSHA deadlines of 8 or 24 hours.

Who should investigate a robot near miss?

A small team: the area supervisor, an EHS representative, and someone who can read the robot’s controller logs, typically a robotics or maintenance engineer. Workers involved in the event should participate as information sources, never as targets. Blame-driven investigations teach people to stop reporting, which is the most expensive outcome of all.

This article is for informational purposes only and does not constitute legal, regulatory or professional safety advice. Check your jurisdiction’s reporting duties with a qualified professional.

Sources

• OSHA, Incident Investigation
• OSHA, 29 CFR Part 1904, Recording and Reporting Occupational Injuries and Illnesses
• OSHA, Report a Fatality or Severe Injury (29 CFR 1904.39)
• OSHA, 29 CFR 1910.147, The Control of Hazardous Energy (Lockout/Tagout)
• NIOSH, Robotics in the Workplace: An Overview
• Layne LA, Robot-related fatalities at work in the United States, 1992-2017, American Journal of Industrial Medicine, 2023
• National Safety Council, Injury Facts: Work Injury Costs