Why Human Error Is Not a Root Cause

A worker pushes the wrong button. A driver fails to see another vehicle. A technician skips a step. A supervisor misses a hazard. An employee forgets to complete an inspection. The investigation begins.

Then someone writes:

Root Cause: Human Error

The investigation is finished. Except it probably is not. That distinction is critical. If your investigation stops at human error, you may be stopping exactly where the most useful part of the investigation should begin.

What Is Human Error?

Human error generally describes an unintended action or decision that differs from what was expected.

People:

• forget,
• misunderstand,
• misread,
• misidentify,
• overlook,
• select the wrong control,
• perform steps out of sequence,
• make incorrect assumptions, and
• sometimes simply make mistakes.

That is not controversial. People make errors.

The problem begins when human error becomes the explanation for the event.

Consider:

The employee selected the wrong switch because of human error.

What have we learned?

Not much.

Now ask:

• Were the switches identical?
• Were they positioned beside each other?
• Were they clearly labeled?
• Was the employee interrupted?
• Was visibility poor?
• Was the task unfamiliar?
• Did the employee receive useful feedback after operating the switch?
• Could the system detect the incorrect selection?
• Could the employee recover before consequences occurred?

Now we are investigating.

Human Error Describes the Action, Not the System

Suppose a forklift operator backs into a fixed object.

The investigation states:

Cause: Operator failed to maintain awareness of surroundings.

That may describe the operator's behavior.

But consider what else could have influenced the event:

• limited visibility,
• equipment blind spots,
• noise,
• pedestrian activity,
• poor lighting,
• congestion,
• time pressure,
• an unexpected obstruction,
• layout,
• alarms,
• mirrors or cameras,
• simultaneous demands, or
• the design of the work area.

The operator's action still matters. But it occurred inside a system. A useful investigation examines both.

DOE's Human-Performance Model Makes This Point Explicit

The U.S. Department of Energy's Human Performance Improvement framework describes human performance as a system involving:

• organizational factors,
• job-site conditions,
• individual behavior, and
• results.

Within that model, human error is not treated as a sufficient explanation by itself. Errors are connected to the tools people use, the tasks they perform, the conditions surrounding the work, and the organizational system supporting them. That gives us a much more useful investigative question.

Not simply:

“Who made the error?”

But:

“What conditions made this error possible or more likely, and what allowed it to produce this consequence?”

The Same Error Can Produce Different Outcomes

Imagine two employees make exactly the same mistake. Employee A selects the wrong control. An interlock prevents the equipment from operating. Nothing happens. Employee B selects the wrong control on a different machine. There is no interlock. Someone is seriously injured. The human error may be essentially identical.

The outcomes are completely different.

Why?

Controls.

This is one reason human error is such a weak stopping point for investigation.

If we focus only on the employee's action, we can miss the systems that determine whether an ordinary mistake becomes:

• harmless,
• recoverable,
• damaging, or
• catastrophic.

Ask What Was Supposed to Prevent the Consequence

When human error appears in an event, ask:

What was supposed to prevent this error from producing harm?

Perhaps:

• an interlock,
• a guard,
• a warning,
• a verification step,
• a spotter,
• a physical barrier,
• separation,
• a redundant check,
• equipment design,
• an alarm, or
• another critical control.

Then ask:

• Was the control present?
• Was it functional?
• Was it independent?
• Could it be bypassed?
• Did employees know how to use it?
• Did conditions reduce its effectiveness?
• Was there a backup if it failed?

Now the investigation is moving beyond error toward system performance.

“They Failed to Follow the Procedure” Is Not Much Better

Another common investigation conclusion is:

Root Cause: Employee failed to follow procedure.

Again, that may describe what happened. But it creates more questions.

Why wasn't the procedure followed?

Possibilities include:

• the procedure did not match the work,
• the employee did not know it existed,
• the procedure was unavailable,
• it was difficult to understand,
• equipment had changed,
• the procedure conflicted with another requirement,
• experienced employees routinely performed the task differently,
• supervisors accepted the alternate method,
• following the procedure created another problem, or
• the employee intentionally chose not to follow a known workable requirement.

Those are very different situations.

Writing procedure not followed does not distinguish among them.

Investigation should.

A Procedure Violation May Be the Beginning of the Investigation

Suppose a written procedure requires five steps. Employees routinely use three.

You could conclude:

Employees are violating the procedure.

Or you could investigate.

Ask:

• When did the three-step method begin?
• Who uses it?
• Do supervisors know?
• Why is it used?
• Has the equipment changed?
• Does the five-step method still work?
• What advantages does the alternate method provide?
• What risks does it introduce?
• Has the three-step method usually succeeded?
• What happened differently during the event?

Now you are learning about the work. The result may still be that employees need to follow the original procedure. Or you may discover that the procedure needs to change. You should know which before deciding.

Human Error Is Normal—But That Does Not Make It Acceptable

Human and Organizational Performance—HOP—commonly begins with the principle:

Error is normal.

This is sometimes misunderstood.

It does not mean:

Errors do not matter.

It means:

Human beings are fallible, so systems should not depend on perfect human performance.

A high-consequence operation should account for the possibility that someone will eventually:

• forget,
• misread,
• misunderstand,
• select incorrectly,
• become distracted, or
• make an incorrect assumption.

That leads to a better design question:

“What happens when a good employee makes an ordinary mistake?”

If the answer is:

“Someone can be killed,”

the organization should examine its controls.

Error-Likely Situations

Errors are not necessarily random. Certain conditions can make them more likely.

DOE's human-performance framework refers to error precursors and error-likely situations.

Examples can include:

• time pressure,
• high workload,
• multiple simultaneous tasks,
• unfamiliar work,
• unclear standards,
• distractions,
• interruptions,
• changes from normal conditions,
• confusing controls,
• unavailable tools,
• poor equipment layout,
• fatigue,
• assumptions, and
• limitations in attention or short-term memory.

This gives investigators something actionable.

Instead of saying:

“The employee made an error.”

ask:

“What made this an error-likely situation?”

Look at the Conditions Surrounding the Action

Imagine a maintenance employee opens the wrong valve.

Human error?

Yes.

But now you discover:

• both valves are identical,
• they are immediately beside each other,
• labels are difficult to read,
• the employee was interrupted,
• the procedure requires switching repeatedly between the document and equipment,
• there is no independent verification, and
• the system gives no immediate indication that the wrong valve was operated.

The event is no longer mysterious. The employee still opened the wrong valve. But now the organization has multiple opportunities for improvement. That is the value of going deeper.

Do Not Replace “Human Error” With Another Label

There is a trap here.

Organizations sometimes stop using the phrase human error but replace it with:

• complacency,
• lack of situational awareness,
• poor attitude,
• carelessness,
• inattention, or
• failure to recognize the hazard.

The investigation has not necessarily improved. It may simply have replaced one label with another.

For example:

Cause: Lack of situational awareness.

Why was awareness lost?

What information was available?

What competing demands existed?

What could the employee see?

What were they paying attention to?

What did they expect?

What normally happened?

What changed?

Keep investigating.

“Be More Careful” Is Not a Strong Corrective Action

If the cause is:

Employee wasn't paying attention

the corrective action often becomes:

Remind employees to pay attention.

Or:

Conduct retraining.

Or:

Tell employees to be more careful.

Those actions may occasionally be appropriate.

But ask:

What changed in the system?

If nothing changed except the employee receiving another reminder, the same conditions may still exist for the next person. Strong corrective actions should address what the investigation actually found.

Retraining Is Not Automatically the Answer

Training solves a knowledge or skill problem.

If the employee:

• did not know the requirement,
• did not understand the hazard, or
• lacked the skill necessary to perform the task,

training may be appropriate.

But training does not repair:

• poor equipment design,
• confusing controls,
• weak guarding,
• inadequate staffing,
• conflicting procedures,
• production pressure,
• missing tools,
• poor layout, or
• ineffective critical controls.

Before recommending retraining, ask:

“Did the person already know what they were supposed to do?”

If the answer is yes, investigate further.

Accountability Still Matters

Moving beyond human error does not mean eliminating accountability. This distinction is essential.

There is a difference between:

• an unintended error,
• a misunderstanding,
• a routine adaptation,
• an at-risk choice,
• knowingly violating a requirement,
• reckless behavior, and
• deliberate harmful conduct.

Intent matters. Context matters. History matters. Organizational contribution matters. An investigation should understand those things before making an accountability decision.

The goal is not:

Never blame the worker.

The goal is:

Do not substitute blame for investigation.

Hindsight Makes Error Look Obvious

After an event, we know the ending. That changes how we interpret everything that happened before it. A warning sign now looks obvious. A decision now looks foolish. A hazard now looks impossible to miss. But the person involved did not know the outcome. They experienced the situation moving forward.

The investigator experiences it backward. That difference matters.

Try asking:

• What did the person know at the time?
• What information was available?
• What information was missing?
• What were they expecting?
• What competing demands existed?
• What had happened successfully before?
• Why did the action make sense in that moment?

Understanding why a decision made sense does not mean agreeing with it. It means reconstructing the event fairly.

Previous Success Can Shape Behavior

Imagine an employee has performed a shortcut 200 times. Nothing bad happened.

The shortcut:

• saves time,
• makes the job easier,
• is used by coworkers,
• has been observed by supervisors, and
• has never previously produced an obvious consequence.

On attempt 201, someone is injured. Afterward, the risk may appear obvious. Beforeward, the employee had 200 successful experiences reinforcing the belief that the method worked. That history is part of the context. If the organization wants to change the behavior, it needs to understand why the behavior became normal.

Ask How the Work Is Normally Done

One of the best ways to move beyond human error is to stop studying only the event. Study normal work.

Ask:

• Is this how the task is usually performed?
• Do other employees do the same thing?
• What normally prevents an incident?
• What makes the job difficult?
• Where do people adapt?
• What controls do they rely on?
• What conditions were different this time?

This helps determine whether you are investigating:

one person's unusual behavior

or:

a normal feature of the system that finally produced an unwanted outcome.

Those require very different responses.

Why Doesn't This Happen More Often?

This is another powerful question.

Suppose an investigation discovers:

• a confusing procedure,
• weak separation,
• an unreliable control, and
• a common workaround.

Those conditions have existed for three years. But there has only been one serious event.

Ask:

“Why doesn't this happen every day?”

Perhaps:

• workers normally compensate,
• another control usually works,
• experienced employees detect the problem,
• conditions rarely align,
• the event required multiple failures, or
• the organization has simply been fortunate.

The answer helps reveal how the system actually creates success.

Look for Multiple Contributing Conditions

Complex events rarely require a single explanatory story.

You may find:

• a human error,
• a procedure problem,
• an equipment issue,
• an ineffective control,
• a communication failure,
• time pressure, and
• an organizational decision

all contributing to the same event.

Do not force those findings into one artificial root cause merely because a form contains one box labeled:

ROOT CAUSE

The purpose of investigation is understanding. Not completing the box.

Does That Mean Root Cause Analysis Is Wrong?

No. Root Cause Analysis can be useful. Organizations and regulators continue to use root-cause terminology. The problem is not the phrase. The problem is stopping too early.

A good systems-oriented Root Cause Analysis can examine:

• contributing factors,
• organizational conditions,
• control failures,
• management systems, and
• multiple underlying causes.

A weak analysis can conclude:

Human error.

The quality of the investigation matters more than the label attached to the method.

Human Error Should Trigger More Questions

When an investigation identifies an error, try this sequence.

1. What did the person do?

Describe the action neutrally.

2. What was expected?

Establish the difference between actual and expected performance.

3. What made sense to the person at the time?

Reconstruct their perspective.

4. What conditions influenced the action?

Examine task, environment, equipment, workload and organizational factors.

5. Was this an error-likely situation?

Look for known error precursors.

6. How is the work normally performed?

Determine whether the action was unusual or normal.

7. What controls should have prevented the consequence?

Identify defenses.

8. How did those controls perform?

Look for missing, weak, bypassed or failed controls.

9. What organizational conditions contributed?

Examine planning, design, staffing, training, supervision, maintenance and priorities.

10. What can we change?

Connect corrective actions directly to supported findings. That is a much stronger investigation.

Human Error and Critical Controls

The issue becomes especially important around serious-injury and fatality exposures.

Imagine a fatal hazard is controlled primarily by:

“Employee remembers to complete Step 7.”

That should make you uncomfortable.

For high-consequence hazards, ask:

• What critical control prevents exposure?
• Is it independent?
• Can its presence be verified?
• Can failure be detected?
• Can an employee recover from an error?
• What happens if the person forgets?
• What happens if the person misunderstands?
• What happens if conditions change?

The more severe the potential consequence, the less comfortable we should be relying entirely on perfect human performance.

Design for Error Tolerance

Strong systems anticipate that mistakes will occur.

Where practical, they create opportunities to:

Prevent

Make the error more difficult.

Detect

Make the error visible.

Recover

Allow correction before harm occurs.

Mitigate

Reduce the consequence if the error reaches the hazard.

Examples can include:

• interlocks,
• physical barriers,
• separation,
• keyed connections,
• independent verification,
• alarms,
• engineered limits,
• clear feedback, and
• fail-safe design.

The objective is not to make people irrelevant. It is to make the system more resilient when people are human.

Human Error Is Data

Instead of treating human error as the end of an investigation, treat it as information.

An error may reveal:

• a confusing interface,
• an unrealistic procedure,
• an error-likely task,
• a weak control,
• an organizational conflict,
• a training gap,
• a communication problem, or
• a mismatch between work-as-imagined and work-as-done.

That makes the error valuable to the organization—if the organization is willing to learn from it.

The Better Question

When someone says:

“The worker made a mistake.”

do not argue with them automatically. They may be correct.

Instead ask:

“What made that mistake possible, and what allowed it to matter?”

That question changes the investigation.

It moves us from:

Who failed?

toward:

How did the system perform?

It moves us from:

Who needs retraining?

toward:

What needs improvement?

And it moves us from:

Human error was the root cause

toward something much more useful:

Human error is where our next questions begin.

For a closer look, read The Five Principles of HOP Explained.

Supporting resources

Related commercial publications are disclosed under Authorship & disclosures below.