Process Safety · API Publication 770

Human Error in the Process Industries

Causes, performance shaping factors, and reduction strategies according to API 770 — A Manager’s Guide to Reducing Human Errors

American Petroleum Institute · March 2001

01

The scale of the problem

Over the past 30 years, the 100 largest accidents at chemical and hydrocarbon processing facilities have severely injured or killed hundreds of people, contaminated the environment, and caused more than $8 billion in property damage alone. The true cost was far higher once business interruption, cleanup, legal fees, fines, and market share losses are factored in.

The common thread in nearly all of these events: human error. In systems with a high degree of hardware redundancy — where equipment failures are well defended against — human errors can account for more than 90 percent of the total system failure probability.

>90%Of system failure probability in high-redundancy facilities
80–85%Of human errors stem from work situation design, not the individual
$8B+In direct property losses from the 100 largest process industry accidents
Key finding: A study of 190 accidents at chemical facilities identified four leading root causes: insufficient knowledge (34%), design errors (32%), procedure errors (24%), and operator errors (16%). A parallel study of petrochemical and refining accidents found equipment and design failures at 41%, operator and maintenance errors at another 41%, and inadequate procedures at 11%.

Root causes across 190 chemical industry accidents

Insufficient knowledge 34% Design errors 32% Procedure errors 24% Operator errors 16%
Insufficient knowledge 34%, design errors 32%, procedure errors 24%, operator errors 16%.

Source: API 770, Section 1.1 — Study of 190 accidents at chemical processing facilities


02

Defining human error

API 770 defines human error as any human action — or failure to act — that exceeds the tolerances set by the system with which the human interacts. Minor variations in how a task is performed are normal and usually inconsequential. Only when a variation crosses a system boundary does it become an error.

The standard distinguishes two fundamental categories: unintentional errors (slips and lapses, committed without prior intent) and intentional deviations (shortcuts the worker believes to be correct or more efficient, but which may exceed system tolerances under certain conditions).

Three patterns of performance variability (API 770, Section 2.2)

Random: high dispersion. Systematic: consistent bias. Sporadic: unpredictable outlier.

Random errors require reducing overall variability. Systematic errors require correcting an underlying bias. Sporadic errors are the most difficult to anticipate and control.


03

Performance shaping factors (PSFs)

Anything in the work environment that affects how a worker performs a task is called a Performance Shaping Factor (PSF) by API 770. PSFs fall into two broad families: internal factors (characteristics the worker brings to the job) and external factors (characteristics of the environment, task, and procedures).

Internal PSFs

  • Training and skill level
  • Practice and experience
  • Knowledge of required performance standards
  • Physical and mental stress
  • Motivation and work attitude
  • Emotional state
  • Physical condition and health
  • Culture and group identity

External PSFs — Situational

  • Temperature, lighting, noise, air quality
  • Work hours and shift rotation
  • Availability of special equipment and tools
  • Staffing levels
  • Organizational structure
  • Authority and communication channels
  • Plant policies

External PSFs — Task & Equipment

  • Written or unwritten procedures
  • Cautions and warnings
  • Information and cognitive load
  • Short- and long-term memory demands
  • Feedback and knowledge of results
  • Control and display design
  • Task criticality and repetitiveness

The operator model

  • Sensing, perception, and discrimination
  • Information processing and decision-making
  • Physical or verbal response
  • Internal and external feedback loops
  • PSFs can disrupt any of these three links

Stress level vs. performance effectiveness — Yerkes-Dodson curve (API 770, Figure 6)

Very low stress: low performance. Optimum stress: peak performance. Extreme stress: sharp collapse.

Source: API 770, Figure 6 — Relationship of stress and performance (based on Yerkes-Dodson, Reference 7)


04

Strategies for improving human performance

API 770 estimates that 80 to 85 percent of human errors arise primarily from the design of the work situation — the tasks, equipment, and environment — which managers directly control. Only 15 to 20 percent are primarily driven by individual human characteristics that are difficult to change.

This finding has a direct management implication: the most effective strategy is to focus on the work situation, not on blaming or disciplining the individual.

Primary origin of human errors — work situation vs. individual characteristics

Work situation design 82% Individual characteristics 18%
82% work situation design, 18% individual characteristics.

Source: API 770, Section 3 — Strategies for improving human performance

The recommended strategy — the work-situation approach — involves five action lines that managers can implement directly:

Human error opportunities across the plant life cycle (API 770, Table 1)

Eight stages from R&D through corporate management with varying relative error exposure.

Source: API 770, Table 1 — Human error opportunities across the plant life cycle


05

Human reliability analysis (HRA)

When a numerical estimate of human error probability is required — for instance, as an input to a quantitative risk assessment — API 770 describes the use of Human Reliability Analysis (HRA). HRA complements good human factors engineering; it does not replace it.

Charter the analysis

Define scope, objectives, resources, responsibilities, timeline, and acceptance criteria before beginning the work.

Select an HRA technique

Choose among methods such as THERP, HEART, or SLIM based on data availability and task complexity.

Apply the technique

Decompose tasks into sub-tasks, identify error modes, estimate probabilities, and quantify PSF effects.

Understand the limitations

HRA estimates carry wide uncertainty ranges. Use them for comparing alternatives — not as absolute probability values — and document all assumptions.


06

Conclusion: from blame to system design

API 770 closes with a fundamental message for process industry managers: disciplining the individual who made an error does not eliminate the root cause. The vast majority of errors are committed by skilled, careful, well-meaning employees working in poorly designed work situations.

The paradigm shift the standard proposes: Instead of looking for someone to blame, identify the work-situation conditions that made the error possible — then correct them systematically. This work-situation approach is what produces lasting improvements in process safety, product quality, and operational productivity.

Involving workers themselves in identifying and designing improvements is not just good practice: it is the factor most likely to determine whether any intervention succeeds or fails. Workers possess the most direct knowledge of the factors that hinder their own performance — and they will support the strategy far more enthusiastically if they are not penalized for telling the truth.