Gemba walks were designed to help leaders see reality on the shop floor. In most organisations, they produce compliance data instead because the walk itself is set up to check whether work matches the standard. It can confirm or deny the procedure, but it cannot surface what the procedure never covered.
Ohno’s original concept was closer to an experiment than an audit. Spear and Bowen (1999) found that Toyota treated its standards as “ideas to be tested” by the people doing the work. When Western lean practice adopted Gemba, it kept the walk but replaced the experiment with a checklist.
Here is what Gemba was built to do, how verification replaced learning, and what changes when leaders ask different questions.
Key findings
- Taiichi Ohno’s genchi genbutsu meant “go and see for yourself.” Western lean practice converted it into “go and check against the standard” (Ohno, 1988).
- Checklist-driven Gemba walks produce less useful safety data than inquiry-based approaches (Xu et al., 2023).
- The quality of leadership questions predicts safety outcomes more strongly than visit count (Luria and Morag, 2012).
- Across 400 cases, 90% of actions first labelled as operator error traced back to workplace conditions (Bitar et al., 2018).
- Adding Learning from Normal Work tools to existing Gemba processes produces a 50% increase in found hazards and a rise from zero to five error traps per task (p < .001).
Where does Gemba come from?
Gemba is a Japanese word meaning “the real place.” In manufacturing, it refers to the shop floor where products are made and value is created. Learning from Normal Work draws on the same principle but extends it from seeing the standard to understanding the conditions that shape how work is actually done.
Taiichi Ohno, the architect of the Toyota Production System, built genchi genbutsu into Toyota’s operating method in the 1950s (Ohno, 1988). The word translates as “go and see for yourself.” Ohno’s purpose was understanding the system rather than confirming compliance. He wanted leaders to see waste and constraints directly, at the place where the work happens.
Masaaki Imai brought the concept to Western management with Gemba Kaizen (Imai, 1997). He described Gemba as the place where problems become visible and where improvement begins. The idea spread through lean manufacturing and then into safety and quality management across every industry.
What Ohno described was closer to an experiment than an inspection. The leader went to the floor with a question, watched the work, and came back with insight about the system. Spear and Bowen (1999) documented that Toyota treated its standards as “hypotheses to be tested” by the workers doing the work. The standard was a starting point for learning.
What happened next reversed the method’s purpose fully.
Ohno’s cycle started with a question about the system. Most Western Gemba walks start with a checklist about the standard.
What assumptions underpin a Gemba walk?
Five assumptions shape the way most organisations run their Gemba walks. Learning from Normal Work tests each one against what workers actually face on the floor.
The procedure describes how work is done. If the standard captures reality, then checking against it is enough. Hollnagel (2014) showed that a gap always exists between work-as-imagined and work-as-done, and that this gap grows as the system gets more complex.
Noncompliance signals a problem with the person. When a leader sees a deviation, the default response is correction. This assumes the worker chose to deviate rather than adapted to a constraint the procedure did not cover.
The leader already knows what to look for. A checklist guides what gets checked, which also controls what gets missed. Anything outside the checklist stays invisible no matter how many walks the leader completes.
Observation alone surfaces the important things. Many conditions cannot be seen by watching. Button labels that have worn off and workarounds built into muscle memory do not show up unless someone asks the right question.
If work matches the standard, the system is safe. This is the base on which the others rest. A clean Gemba report confirms compliance, which leaves every hazard that lives outside the standard unnoticed. Deming showed that 94% of quality problems belong to the system rather than to the person doing the work (Deming, 1986). When the walk checks only the standard, 94% of the problem space stays out of view.
What were Gemba walks built to achieve?
Gemba walks were built to close the distance between leaders and the work they manage. Learning from Normal Work pursues the same goal through methods that surface what observation alone cannot.
Ohno wanted leaders to watch the flow of goods and data at the point where value is created (Ohno, 1988). His goal was to see waste with his own eyes rather than reading about it in a report. The output was a deeper grasp of the system, never a form with tick boxes.
Imai (1997) gave Gemba two core jobs: keeping the standard and making things better. Both jobs require the leader to learn something new during the visit. A walk that only confirms what was already known does the first job and misses the second.
In lean practice, the intended outputs include:
- Direct view of flow, waiting, rework, and bottlenecks
- Conversations with workers about what makes the job difficult
- Improvement ideas that come from the floor rather than from management
- Leadership presence that signals respect for the work
The further the practice drifts from these outputs, the less value it produces. When the main output becomes a form with compliance scores, the walk has shifted from learning to checking. The leader returns with a number that says the floor is in order, but the number cannot show what the form did not ask about.
The gap between what Gemba was built to produce and what most walks actually deliver.
How did most organisations turn Gemba walks into audits?
The shift happened when lean met compliance. Learning from Normal Work reverses this shift by replacing the checking mindset with structured questions about real work.
Western firms adopted Gemba walks as part of lean rollouts, but they added things Ohno never included. Checklists made the walk the same every time. Scoring systems turned findings into metrics. Safety and quality teams merged Gemba with their existing audit programmes, adding compliance points to every visit.
Luria and Morag (2012) found that the quality of leadership questions predicts safety outcomes more strongly than visit count. Open questions produced useful findings, while closed questions confirmed what the leader already believed. Most Gemba programmes track the number of walks completed per month rather than what the walks actually surface.
Spear and Bowen (1999) described Toyota’s approach differently. Standards were ideas to be tested by the workers doing the work, but Western lean practice reversed this. The standard became a rule, and the Gemba walk became the tool for checking whether people followed it. The walk kept its name but changed its function.
Leaders who run checking walks are doing what the programme asks of them. The programme itself has moved from what Ohno built to something closer to an audit with a different name. The intent was learning, but the system rewards counting.
A checklist sets the boundary of what a walk can find. Everything outside that boundary stays invisible.
What are the limits of verification as a Gemba strategy?
Checking finds what it was built to find: whether work matches the standard. Learning from Normal Work starts from a simple truth: what you look for is what you find, and what you do not look for stays hidden.
Xu et al. (2023) showed that checklist-driven Gemba walks produce less useful safety data than inquiry-based approaches. The checklist defines the boundary of what the walk can see. Everything outside that boundary remains hidden, no matter how thorough the checker is.
I ran a Walk-Through Talk-Through at an industrial site where crane operators showed me their remote controls. The procedure said “operate the remote control to position the load.” What the operators actually faced was different:
- Buttons stuck in the pressed position, forcing operators to pry them free mid-lift
- Batteries died without warning while loads hung overhead
- Labels had worn off, so operators worked from memory of button positions
- Several operators had opened the remote casings and rewired the controls to make them work again
A verification walk would have checked whether the remote was used. It would not have asked what makes operating it difficult.
Workers adapt to constraints the procedure does not cover. When a leader arrives with a form, the worker shows the standard version of the task. The shortcuts that keep output flowing stay hidden, because they are the very things the form would flag.
This is the core limit of any checking-based approach. The walk can tell you whether the standard was met. It cannot tell you whether the standard describes what actually happens between one visit and the next.
Sexton et al. (2018) found that leadership walkrounds produce safety culture gains only when leaders give feedback on what they learn. A walk that produces only compliance data has nothing new to feed back, because it measured the standard rather than the work.
How does Learning from Normal Work change what leaders find?
Learning from Normal Work changes what leaders find because it changes what they look for. Rather than checking if work matches the standard, leaders ask what makes each step of the task hard in the real conditions workers face.
The shift starts with how noncompliance is read. Checking treats a shortcut as a fault of character or focus. LFNW treats it as a response to a constraint the system created. When a worker skips a step, the question changes from “why didn’t you follow the procedure?” to “what made that step difficult?” This reframing, grounded in Rasmussen’s (1997) work on system drift, reveals conditions that verification cannot see.
The questions matter as much as the mindset. PATH Dialogue gives leaders structured questions drawn from applied psychology and task analysis research, put into language any supervisor can use on the floor. The questions are designed to find what the worker knows but the system has not yet asked for. A detailed explanation of how PATH Dialogue turns checking walks into learning conversations covers the method step by step.
Walk-Through Talk-Through extends this further by breaking the task into steps and walking through each one with the worker. PATH Dialogue works at the talk level. WTTT works at the task level, turning up the error traps hidden in each step. These tools give leaders a structured way to find what checking-mode Gemba was never built to see.
Data from BP shows what this approach reveals. Across 400 cases, 90% of actions first labelled as operator error traced back to workplace conditions (Bitar et al., 2018). Adding LFNW tools to existing Gemba processes produces a 50% increase in found hazards and a rise from zero to five error traps per task (p < .001).
A pharma company, a major lean practitioner, added Learning from Normal Work to its existing Gemba walks and risk reviews (Psychology Applied, 2023). The addition works because the Learning from Normal Work methodology builds on what is already in place. LFNW extends Gemba by restoring the testing spirit that Ohno intended, giving leaders the questions and tools that verification-mode practice removed.
The result is better questions on the walks leaders already do, producing the insight that Ohno went to the floor to find.
Three changes turn a verification walk into a discovery conversation: what you look for, how you ask, and how you interpret what you find.
References
Deming, W. E. (1986). Out of the Crisis. MIT Press.
Hollnagel, E. (2014). Safety-I and Safety-II: The Past and Future of Safety Management. Ashgate Publishing.
Imai, M. (1997). Gemba Kaizen: A Commonsense Approach to a Continuous Improvement Strategy. McGraw-Hill.
Luria, G. & Morag, I. (2012). Safety management by walking around (SMBWA): A safety intervention program based on both peer and manager participation. Accident Analysis & Prevention, 45, 248-257.
Ohno, T. (1988). Toyota Production System: Beyond Large-Scale Production. Productivity Press.
Psychology Applied Ltd. (2023). Learning from Normal Work: Enterprise Implementation Brochure. Psychology Applied.
Rasmussen, J. (1997). Risk management in a dynamic society: A modelling problem. Safety Science, 27(2-3), 183-213.
Sexton, J. B. et al. (2018). Providing feedback following Leadership WalkRounds is associated with better patient safety culture, higher employee engagement and lower burnout. BMJ Quality & Safety, 27(4), 261-270.
Spear, S. & Bowen, H. K. (1999). Decoding the DNA of the Toyota Production System. Harvard Business Review, 77(5), 96-106.
Xu, J. et al. (2023). Effectiveness of safety leadership walkrounds: A systematic review. Journal of Safety Research, 84, 280-300.
