A study of 1.3 million BBS observations across 88 industrial sites found that the sites pushing their observation rates highest saw the smallest gains. What predicted results was the effort each observer put in (Spigener, Lyon and McSween, 2022). Human and Organizational Performance (HOP) starts from a different premise: behavior is shaped by system conditions, not by choice.
That shift changes what a field visit can find.
Below, I compare the two methods and set out a practical upgrade path.
Key findings
- Behavior-Based Safety reliably raises observed safe behavior percentages, but the largest dataset (1.3 million observations across 88 international clients) found no link with incident reduction.
- W. Edwards Deming estimated that 94% of performance variation belongs to the system, leaving BBS programs addressing about 6% of causal factors.
- Human and Organizational Performance (HOP) treats behavior as a symptom of system conditions rather than the root cause of incidents.
- At one of our manufacturing customers, eight BBS observation cards at a single workstation produced zero improvement suggestions, while one Learning from Normal Work session surfaced multiple error traps and worker-generated solutions.
- Organizations can upgrade existing BBS programs by retraining observers to ask open questions about constraints rather than check compliance against checklists.
What is Behavior-Based Safety?
Behavior-Based Safety (BBS) grew out of applied psychology in the 1970s. If unsafe behaviors cause incidents, then observing and correcting those behaviors should reduce them.
In practice, a BBS program puts trained observers in the field to watch workers and score them against a checklist. Observers give feedback on the spot, and the data is tracked as safe behavior scores over time.
BBS programs became common in high-hazard industries during the 1980s and 1990s. They brought in a structured reason for leaders to be in the field, talking to workers about how tasks get done.
That field presence is worth keeping.
The challenge is what the checklist picks up. It checks listed behaviors and records only what the observer expects to see. Risks the observer does not know to look for stay hidden from the checklist.
Learning from Normal Work fills this gap with structured questions that surface constraints, as this guide to observations and learning conversations shows.
Seeing where BBS reaches its limits makes it clear what Human and Organizational Performance (HOP) adds.
What is Human and Organizational Performance?
Human and Organizational Performance (HOP) is built on five principles: people make errors, blame fixes nothing, context drives behavior, learning is vital, and how leaders respond matters.
These principles shift focus from the individual to the system conditions that shaped what they did.
Where BBS asks “did the worker follow the procedure?”, HOP asks what conditions made the task hard and what the worker had to adjust to. This moves the worker’s role from watched subject to expert source.
Hollnagel’s Safety-II framework (2014) showed that everyday adaptations keep operations running safely, not just the absence of failures. These adaptations are invisible to behavioral checklists, which is why observation alone cannot capture them.
Reason’s Swiss Cheese Model (1997) split active failures at the frontline from the latent conditions that set the stage for error. HOP turns these models into practical questions leaders can ask in the field.
Learning from Normal Work is a method built by Psychology Applied that puts HOP into practice through 14 structured tools. Walk-Through Talk-Throughs and Learning Teams surface what makes the work hard while there is still time to act.
Both methods aim to reduce incidents, but they differ on what causes them.
How do BBS and HOP differ?
The core split is about cause. BBS treats behavior as the source of harm; HOP treats it as a symptom of system conditions.
Behavior-Based Safety (BBS) and Human and Organizational Performance (HOP) split along the line that Lee Ross (1977) called the fundamental attribution error. People tend to blame the individual and miss the role of situational factors.
When a boss sees a worker take a shortcut, the first reason they reach for is lack of care.
But the real cause often lies elsewhere: the procedure may assume time the schedule does not allow, or the right tool may have been missing on that shift.
Across 400 cases, 90% of behaviors initially labeled as unsafe traced back to workplace conditions (Psychology Applied, 2022-2026). Deming (1993) put the split at 94% system to 6% individual.
No matter the exact split, BBS puts most of its focus on the smaller share of what drives incidents.
One leader summed it up during a Learning from Normal Work rollout: staff kept hearing about the rules, but no one asked what made them hard to follow.
| Dimension | BBS | HOP / Learning from Normal Work |
|---|---|---|
| Unit of analysis | Individual behavior | System conditions |
| Primary method | Observation against a checklist | Structured inquiry with workers |
| Core belief | Behavior causes incidents | System conditions shape behavior |
| Data produced | Safe/unsafe behavior scores | Constraints and error traps |
| Worker role | Watched subject | Expert source |
| Leader role | Checker | Explorer and learner |
| Addresses | About 6% of causal factors | About 94% of causal factors |
| Ceiling | Plateaus at compliance level | Keeps improving through system redesign |
Deming put 94% of performance variation down to the system, leaving behavioral programs working on the last 6%.
Does the evidence back one method over the other?
What does the evidence say about BBS results?
The data tests the belief that Behavior-Based Safety (BBS) programs bring lasting safety gains.
Deming (1993) said that 94% of performance variation belongs to the system and only 6% to the individual. Even a perfect BBS program at full strength would cover about 6% of what drives incidents.
Spigener, Lyon and McSween (2022) looked at 1.3 million observations from 88 international sites. Injuries fell by 42% over three years, so the programs clearly did something. But the sites that pushed observation rates highest per employee saw smaller gains than those in the middle range. The best results belonged to observers who each put real effort into their own observations.
Carra and others (2024) reviewed about 230 research papers on BBS from the 1970s to 2023. Most measured behavior change rather than injury reduction, and the strongest designs showed the weakest effects on actual safety outcomes.
| Evidence source | Key finding |
|---|---|
| Deming (1993) | 94% of variation is systemic; BBS covers about 6% |
| Spigener, Lyon & McSween (2022) | 1.3 million observations: results tracked each observer’s effort rather than the site’s observation rate |
| Carra et al. (2024) | Strongest designs show weakest effects on safety outcomes |
From a Human and Organizational Performance (HOP) view, these findings confirm that a program focused on individual behavior reaches a ceiling no matter how well it is run.
The BP Texas City explosion in 2005 shows what these numbers mean in practice (Hopkins, 2008). Safe behavior scores at the site were rising while process safety was getting worse.
Learning from Normal Work, asking workers what they adapted to during routine startups, could have surfaced these risks. Five years later, the Deepwater Horizon showed the same pattern: an active BBS program coexisted with deadly systemic risk.
If observation volume alone does not reduce incidents, what does a field visit find when the method changes?
What does HOP discover that BBS cannot?
Human and Organizational Performance (HOP) finds what existing systems miss, while Behavior-Based Safety (BBS) checks what is already known.
At one of our customers, a BBS program had been running for a few years when returns started to diminish. Observers were auditing rather than learning, and the findings were getting thinner.
Eight observation cards filled in at a single workstation gave zero improvement suggestions.
When we used a Walk-Through Talk-Through from Learning from Normal Work at the same spot, the findings came at once.
- The machine’s safety guard could trap the worker’s hands when it came down.
- Cutting tools could be put in the wrong way round due to their shape.
- The worker had built a way to handle varying roll weights that was in no written procedure.
- Design fixes came from the worker: a poka-yoke so tools could only go in one way, and limiters on the guard to stop the trapping risk.
The worker is the expert on how the task gets done. LFNW gives the tools that make this kind of finding happen each time, like learning conversations that replace standard BBS cards.
Eight observation cards at one workstation gave zero improvement suggestions; a single Walk-Through Talk-Through surfaced four.
For groups already running BBS programs, the practical question is how to build on them.
Can BBS and HOP coexist?
Human and Organizational Performance (HOP) builds on what Behavior-Based Safety (BBS) has already created, so the two can coexist. The upgrade path works better than a fresh start.
What BBS built in the field is worth keeping. What changes is the focus: from scoring behavior against a checklist to asking what makes the work hard.
At one of our customers, we redesigned a BBS program into a two-tier system.
- Level 1 kept its observers but gave them open questioning from Learning from Normal Work, replacing the checklist with conversations that explore constraints.
- Level 2 added planned Walk-Through Talk-Throughs and Learning Teams for deeper systemic analysis.
The group kept its observers but gave them better questions and a different purpose.
This is the shift that observer training needs to make: from teaching observers to check and judge, to teaching them to ask and listen. Knowing work as done versus work as imagined becomes the foundation for every field conversation.
Learning from Normal Work gives the structured pathway to evolve existing programs into systems that surface the conditions shaping behavior rather than simply measure compliance.
Upgrading keeps existing observers at Level 1 and adds deeper study at Level 2.
References
Carra, S., Bottani, E., Vignali, G., Madonna, M., & Monica, L. (2024). Implementation of behavior-based safety in the workplace: A review of conceptual and empirical literature. Sustainability, 16(23), 10195.
Dekker, S. (2014). The Field Guide to Understanding ‘Human Error’ (3rd ed.). CRC Press.
Deming, W.E. (1993). The New Economics for Industry, Government, Education. MIT Press.
Geller, E.S. (2005). Behavior-based safety and occupational risk management. Behavior Modification, 29(3), 539-561.
Hollnagel, E. (2014). Safety-I and Safety-II: The Past and Future of Safety Management. CRC Press.
Hopkins, A. (2008). Failure to Learn: The BP Texas City Refinery Disaster. CCH Australia.
Spigener, J., Lyon, G. & McSween, T. (2022). Behavior-based safety 2022: Today’s evidence. Journal of Organizational Behavior Management, 42(4), 336-359.
Reason, J. (1997). Managing the Risks of Organizational Accidents. Ashgate.
Ross, L. (1977). The intuitive psychologist and his shortcomings: Distortions in the attribution process. In L. Berkowitz (Ed.), Advances in Experimental Social Psychology, 10, 173-220.
Sulzer-Azaroff, B. & Austin, J. (2000). Does BBS work? Behavior-based safety and injury reduction: A survey of the evidence. Professional Safety, 45(7), 19-24.
Data source: Psychology Applied measurements across enterprise clients (2022-2026).
