Line down. Four hours of unplanned downtime on a Sunday morning, costing us production time against an automotive customer that measures delivery tolerance in minutes, not hours.
The process engineer told me it was the exact same failure as the previous week: a thermistor in heating zone three. When I asked why it had failed again after he supposedly fixed it, he shrugged. "I repaired it, but clearly not definitively."
That word—definitively—is the chronic disease of manufacturing quality. Over twenty years managing ISO 9001 and IATF 16949 systems, I have lost count of how many times we applied a patch, only for the same failure to return a month later. We treat symptoms because treating symptoms is fast. We wait for problems to hit us, rather than going to find them.
I decided to stop waiting. The next time a chronic issue appeared, I would go to it. This is how I discovered the rigorous discipline of Jishuken.
Jishuken vs. Standard Kaizen
In the Toyota Production System, Jishuken (自主研) translates directly to voluntary or self-study. In practice, it means management-led deep investigation. It is not a delegated team task, nor is it a desk audit.
Standard Kaizen asks operators to make incremental daily improvements. Jishuken demands that managers roll up their sleeves, stand at the machine, and drive fundamental root cause analysis themselves.
A Kaizen Blitz asks: "How do we do this slightly better or faster?" Jishuken asks: "Why are we doing this way at all, and what is the physical reality causing this defect?"
Approaches to Process Improvement
Standard Problem Solving
- Reliance on operator-reported defects and automated alarms
- Data reviewed in meeting rooms or via automated dashboards
- Solutions implemented via lengthy capital approval chains
- Root cause often generalised to 'operator error' or 'machine wear'
Jishuken (Management-Led Study)
- Direct physical observation of the process during the defect
- Variables measured in situ alongside standard production data
- Immediate physical prototyping and iteration on the floor
- Root cause traced to specific environmental or systemic physics
The Five Pillars of Management-Led Investigation
To execute Jishuken correctly, you must abandon the idea of a quick assessment. It requires a structured descent into the operational reality.
1. Sustained Personal Observation
This is not a standard gemba walk. A manager conducting Jishuken spends days at the process. When investigating the thermistor issue, I spent three days at heating zone three.
I mapped temperature profiles against historical downtime logs. The automated reports blamed the sensor. Direct observation revealed the truth: the temperature regulation reacted to sudden air pressure drops in the hall whenever the loading bay doors opened.
2. Relentless Root Cause Analysis
You must push past the first three levels of 'Why?' Tools like 5 Why, Ishikawa, and PFMEA are vehicles, not destinations. The goal is to understand the physical mechanism causing the failure.
Because the sensor sat in a draught, the solution was not a recalibration. The solution was relocating the sensor and installing an aerodynamic baffle between the loading bay and the production hall.
3. Extracting Tacit Operator Knowledge
Operators hold critical process knowledge they rarely write in standard work documents. An operator on that line had 22 years of experience. He casually mentioned that the temperature jumped whenever the wind blew from the hangar. No previous investigation had captured this detail, because no one had asked him correctly or in the right context.
4. Immediate Experimentation
Jishuken ends with validation on the floor, not in a future quarterly budget cycle. When we installed the physical baffle, thermal stability improved immediately. Temperature-related stoppages dropped to zero from that shift forward.
5. Yokoten (Standardised Deployment)
If a fix only exists on one line, you have only solved a local symptom. We shared the baffle design and sensor relocation procedure across all five plants in the group via Yokoten. Two other plants discovered the identical draught issue and implemented the fix immediately, preventing future downtime.
Executing a Jishuken Cycle
A proper Jishuken cycle takes five weeks. It targets one chronic, recurring issue that standard 8D methodologies have failed to eliminate permanently. The criteria for selection are strict: the problem must have recurred at least three times in the last quarter, previous fixes must have only yielded temporary relief, and the financial or quality impact must be clearly measurable.
The Five-Week Jishuken Framework
- 01Week 1: Theme SelectionTarget a chronic, months-old problem that everyone accepts as unavoidable, not an urgent daily crisis.
- 02Week 2: Gemba ObservationManager spends two full shifts on the floor with a stopwatch, tracking environmental and process variables.
- 03Week 3: Causal AnalysisHypothesis testing using multi-vari studies or regression analysis, validated directly against the physical line.
- 04Week 4: ImplementationDeploy a robust, simple countermeasure immediately. Prototype and iterate in real-time.
- 05Week 5: StandardisationVerify the data, lock the new standard work, and deploy the solution to identical processes globally.
Why Jishuken Efforts Fail
Most Jishuken attempts collapse due to management behaviour. If you approach the line with an auditor's mindset—hunting for who made a mistake—you destroy trust before the observation begins.
Delegating the study to a team leader is the most common failure mode. Without the manager's direct presence, the investigation loses authority and the team misses the systemic connections that only emerge through sustained leadership observation.
Jumping to a solution is equally fatal. Declaring 'I already know what the problem is' before spending time at the machine guarantees you will treat a symptom. Jishuken demands the patience to let the data and the physical environment speak.
Data tells you a parameter fluctuated; only observation at the source tells you the bay doors caused it.
Data Needs Context
In an era of Industry 4.0, digital twins, and predictive analytics, some argue that physical floor-walking is obsolete. This is a dangerous misconception. The more we rely on automated data, the more we need humans to interpret that data against operational reality.
An IoT sensor will faithfully report that the oven temperature in zone three is unstable. It will flag an alarm. It will trigger an 8D. But the sensor cannot tell you why the instability happens, because the sensor does not know the loading bay door exists.
Machine learning can identify hidden patterns in your scrap rates. Digital twins can simulate process changes before you implement them. But the critical decision—determining whether the mathematical model matches the physical reality on the shop floor—requires a manager who has stood at the machine and watched it run.
The Leadership Requirement
Implementing ISO 9001 systems at plants like a major aerospace manufacturer and SNOP taught me that systemic improvement requires management accountability. Jishuken forces this accountability.
When you personally lead the investigation, you cannot blame the operator, the maintenance team, or the supplier when a fix fails. You analysed the process. You designed the countermeasure. This level of personal responsibility is precisely what shifts a plant's culture from passive reporting to active quality engineering.
The biggest barriers to manufacturing excellence are rarely technical. They are human. Jishuken works because it forces managers to close their laptops, step onto the floor, and find the truth alongside the people who build the product.