PATH talks surfaced 95.7% of error traps at a European factory site while 262 Gemba walks at the same site found 37% (Psychology Applied data, 2026). The gap between watching and learning comes down to question design.
Both methods were used at the same site with the same people. The gap lies in the belief each question carries about where useful knowledge sits.
Key findings
- At a European factory site, PATH surfaced 95.7% error traps while Gemba walks found 37% with the same workforce.
- A review of 31 studies and 45,673 risk calls confirmed that structured tools clearly beat loose expert judgment (Viljoen et al., 2025).
- Question quality at TE Connectivity improved from 58% to 91% after a four-day Learning from Normal Work program (p < .001).
- In a forced-choice study, 86% of trained workers chose PATH for quick insight into a task, while 93% chose Walk-Through Talk-Through for deep grasp of how work happens.
- Nearly a quarter (24.8%) of Gemba walk findings were too vague to code, compared with only 4.3% from PATH talks.
The answer starts with the question itself.
Why does the question you ask decide what you find?
Every question carries a belief about who holds useful knowledge, and that belief shapes what the other person shares.
HOP principles hold that workers make sensible choices based on the conditions they face. Questions built on this idea surface the constraints that shape how work actually happens.
In the Learning from Normal Work book, a director joins workers at lunch and asks “Why do you sometimes cut corners?” Workers respond with guarded silence because the question places them as the cause (Nazaruk, 2025).
At the same company, a supervisor trained in TEDS questioning (Tell, Explain, Describe, Show) asks “Describe what makes this step difficult.” Within 15 minutes, he surfaces constraints that months of director visits had missed.
This pattern also applies in group settings. A review of three decades of hidden-profile research found that groups spend more discussion time on what several members already know (Sohrab, Waller and Kaplan, 2015).
What only one person knows gets less air time in group choices. In a Learning from Normal Work context, a workaround known to one worker will not emerge unless the guide asks for it directly.
For site HSE managers, swapping the method without changing the question tends to produce the same findings in a new format.
Data from a single site shows just how wide this gap becomes in practice.
What did 262 Gemba walks reveal compared with structured talks?
Structured talks produced different findings from Gemba walks at the same European factory site. A content review of 262 Gemba walks alongside PATH talks and Walk-Through Talk-Throughs showed clear gaps in what each method found (Psychology Applied data, 2023).
| Method | Error traps surfaced | Findings too vague to code | Primary finding type |
|---|---|---|---|
| PATH | 95.7% | 4.3% | Upstream system constraints |
| Walk-Through Talk-Through | 83.9% | Not split out | Task-level constraints |
| Gemba walk | 37.0% | 24.8% | Surface-level hazards (34.7%) |
Three patterns stand out from this Learning from Normal Work data:
- PATH captures upstream system constraints such as kit design flaws and workload pressure that checklists miss.
- Nearly a quarter (24.8%) of Gemba walk findings were too vague to code, compared with only 4.3% from PATH.
- Structured questions improved both the type and the quality of findings.
A pre-set review of 31 studies covering 45,673 risk calls backed this finding across fields (Viljoen et al., 2025). Structured tools clearly beat loose expert judgment. Upstream conditions such as workload pressure and design clashes stay largely hidden from the Gemba method.
Three methods, three discovery profiles, one site.
The practical question for site managers is which method to deploy and when.
How do these methods compare for site managers?
Each method has its own finding pattern that serves a different learning role within a multi-tool proactive learning approach. The table below maps four methods across what matters to site HSE managers.
| Factor | Gemba walk | PATH | Walk-Through Talk-Through | Learning team |
|---|---|---|---|---|
| What it surfaces | Surface hazards, tidiness, rule gaps | Upstream system constraints and error traps (95.7%) | Task-level constraints and error traps (83.9%) | Work-as-done, how people adapt, cross-team links |
| Time needed | 15-30 minutes per walk | 15-20 minutes per talk | 60-90 minutes per task review | 2-4 hours per session |
| How often it can run | Daily to weekly | Weekly to fortnightly | Monthly | Quarterly at best |
| Guide skill needed | Low: checklist | Moderate: structured question set | High: task review and TEDS questioning | Very high: group guiding, bias handling |
| Data quality | 24.8% too vague to code | 4.3% too vague to code | Structured step-by-step format | Narrative, needs coding setup |
| Best used for | Rule checks, hazard watching | Quick insight into task constraints | Deep grasp of how a task works | Complex problems needing diverse views |
How often a tool runs matters because it decides how fast new constraints reach leaders. Learning from Normal Work places learning teams as a deep review method, but they require pulling several workers off the line for long sessions, which limits how often they can run.
PATH talks can be fitted into routine leadership visits at weekly or fortnightly gaps, giving steady coverage that quarterly sessions cannot match.
Each method serves a different learning function at a different frequency.
Even a well-designed tool tends to produce thin results without the right guide.
Why does guide skill matter more than the method?
A guide who believes rule-breaking causes accidents will ask blame-focused questions no matter which template they use. The format only changes what gets found when the guide’s mindset backs learning-focused questions.
At one site, a supervisor got a Walk-Through Talk-Through template without the matching mindset training. The finished form contained hazard labels and judgments such as “should pay more attention” (Psychology Applied data, 2026).
Because the supervisor believed rule-breaking causes accidents, the structured format did not change what he found.
At TE Connectivity, question quality improved from 58% to 91% after a four-day Learning from Normal Work program (p < .001). Guides shifted from blame-focused to learning-focused questions such as “Describe what makes this step difficult” (Psychology Applied data, 2023).
| Metric | Before program | After program | Result |
|---|---|---|---|
| Question quality score | 58% | 91% | p < .001 |
A 2017 review of process hazard methods confirmed that even well-known structured methods depend on the knowledge and lived experience of the people using them (Cameron et al., 2017). The Learning from Normal Work method tackles this through a guide growth path that builds mindset and hands-on skill before rolling out any tool.
The skill needed shapes when each method works best.
When should you use each method?
Forced-choice data from Psychology Applied research supports a layered approach where each method serves a clear purpose (Psychology Applied research data, 2023).
- Gemba walks handle daily to weekly rule checks and hazard spotting.
- PATH talks offer weekly to fortnightly insight into task constraints during routine visits.
- Walk-Through Talk-Through delivers monthly depth through step-by-step task review.
- Learning teams bring diverse views to problems that span several functions or teams.
| Method | Best for | Workers who chose it |
|---|---|---|
| PATH | Quick insight into a task | 86% |
| Walk-Through Talk-Through | Deep grasp of how work happens | 93% |
Workers who have used both tools see that each serves a different learning role. Using them together works better than picking one over the other.
Learning from Normal Work builds this layered skill by adding structured questions and guide training to the visits already happening. Sites keep their Gemba walks and gain the deeper talks that surface upstream system conditions.
PATH handles the frequency need because a 15-minute talk fits inside a leadership visit already on the calendar. For depth, Walk-Through Talk-Through gives a 60-90 minute step-by-step format.
Layer methods by purpose for broader coverage.
Choosing the right methods is only part of the picture.
What makes a multi-method program last?
Having several ways to talk with workers is a starting point. Two added factors shape whether a Learning from Normal Work program produces lasting gains.
- Guide mindset: when guides believe rule-breaking causes incidents, their questions carry subtle blame no matter the template. LFNW tackles this through training that reframes how leaders see the link between conditions and actions.
- Company follow-through: in program reviews, the lowest-rated area has been turning findings into clear actions (Psychology Applied research data, 2023). Without a way to act on findings, talks produce unused insight.
Time pressure and lack of felt importance prevent companies from turning knowledge into lasting change (Drupsteen and Hasle, 2014). The effort put into better talks goes to waste unless the company has a channel for acting on what it learns.
The Learning from Normal Work method includes a maturity check that helps site managers map which learning functions their current program covers. Knowing where you stand is the first step toward building a system where every talk with a worker adds to what the company knows.
References
Cameron, I., Mannan, S., Neron, E., Piccinini, N., Lees, F., & Theophilus, S. (2017). Process hazard analysis, hazard identification and scenario definition: Are the conventional tools sufficient, or should and can we do much better? Process Safety and Environmental Protection, 110, 53-70. https://doi.org/10.1016/j.psep.2017.01.025
Drupsteen, L., & Hasle, P. (2014). Why do organizations not learn from incidents? Bottlenecks and solutions. Accident Analysis & Prevention, 62, 85-93.
Nazaruk, M. (2025). Learning from Normal Work: How to Reduce Risk When Nothing Goes Wrong. Psychology Applied.
Sohrab, S. G., Waller, M. J., & Kaplan, S. (2015). Exploring the hidden-profile paradigm: A literature review and analysis. Small Group Research, 46(5), 489-535. https://doi.org/10.1177/1046496415599068
Viljoen, S., et al. (2025). Are risk assessment tools more accurate than unstructured judgments? A meta-analysis. Behavioral Sciences & the Law, 43(1), 75-113. https://doi.org/10.1002/bsl.2698
Psychology Applied implementation data (2022-2026). PATH vs EHS Gemba vs WTTT output comparison: content analysis of 262 Gemba walks, 10 PATH conversations, 9 Walk-Through Talk-Throughs at a European manufacturing site.
Psychology Applied research data (2023). PATH and WTTT perception of value: forced-choice comparison, n=14 trained practitioners.
Psychology Applied. (2026). TE Connectivity LFNW case study.
