Measure takt time before any other lean intervention
Rationale: All lean tools (kanban, line balancing, heijunka) depend on an accurate takt time. Without this measurement, any intervention is a local optimisation without reference to the customer.
Do: Calculate takt time with real demand data (rolling quarterly or semi-annual average, not peaks) and actual available time (minus breaks, shifts, planned maintenance). Recalculate when demand varies by more than 15%.
Don't: Use the original line design takt time without verifying whether demand has changed. Do not confuse takt time with cycle time or lead time.
Apply Value Stream Mapping before intervening on a single process
Rationale: Optimising a single process without seeing the overall flow can worsen the system: you increase the speed of the non-bottleneck, increase downstream WIP, and do not reduce the lead time perceived by the customer.
Do: Map the flow from customer order to delivery (current state), identify the bottleneck and scheduling point, then design the future state. Start implementation from the future state, not from local optimisation of the current state.
Don't: Run kaizen events on an island without analysing the upstream and downstream impact in the value stream.
Implement 5S before any automation or new equipment
Rationale: Automating a chaotic process produces automated chaos. 5S eliminates visible abnormalities, creates the operational discipline needed for kaizen, and reduces search time for tools and materials, which is pure muda.
Do: Run 5S as a prerequisite for every new improvement project. Include in the 5S audit not only physical order but also digital data management (shared folders, obsolete documents on servers).
Don't: Do a one-time "spring cleaning" without the fourth and fifth S (standardise and sustain). 5S not maintained regresses within weeks.
Use root-cause analysis (5 Whys) before every corrective action
Rationale: Corrective actions on symptoms are expensive and temporary. The 5 Whys lead to the systemic root cause that, once removed, prevents recurrence.
Do: Document each "why" iteration with verifiable evidence (not opinions). When the answer to the fifth "why" points to a system, procedure, or inadequate training, you are reaching the root cause. Include 5S in the analysis: many problems are rooted in workplace organisation.
Don't: Stop at the third "why" when the answer is "the operator made a mistake". Human error is often a symptom, not the root cause.
Calculate and monitor OEE by component, not just the aggregate value
Rationale: An aggregate OEE of 85% can hide an Availability of 70% compensated by high Performance. Acting on aggregate OEE without separating the three components leads to ineffective interventions.
Do: Track Availability, Performance, and Quality separately. Identify which of the three is the primary loss. Implement targeted actions: maintenance for Availability, time study for Performance, SPC for Quality.
Don't: Use OEE as the only production KPI without contextualising it with asset type, product mix, and required takt time.
Install a kanban system before reducing safety stock levels
Rationale: Reducing stock without a pull replenishment system causes stockouts. Kanban ensures that every withdrawal triggers a replenishment order calibrated to real consumption.
Do: Size kanban with the correct formula: N kanbans = (Demand during lead time + Safety Stock) / Quantity per container. Revise the number of kanbans periodically when demand or lead time changes.
Don't: Cut inventory to improve the balance sheet without a functioning pull system in place. Inventory reduction is a lean result, not the starting point.
Involve operators in the analysis and improvement of their own processes
Rationale: Operators know their process better than anyone else. Improvements imposed from above without their involvement are often disregarded or unconsciously undermined. Operator engagement is the primary success factor for lean programmes.
Do: Structure kaizen events always including line operators on the team. Train operators in basic tools (5S, kanban, 5 Whys) and assign them improvement responsibilities in their own workstation. Celebrate improvements publicly.
Don't: Implement lean as an industrial engineering project managed exclusively by specialists. Lean is a cultural transformation.
Standardise the process before improving it (Standard Work)
Rationale: It is not possible to improve an unstandardised process: every operator executes it differently, making it impossible to measure and replicate results. Standard work is the starting point for continuous improvement, not an obstacle to it.
Do: Document the standard operating method (sequence, cycle time, standard WIP, safety points) for every operation. Train operators on the standard and revise the standard when a better method is found.
Don't: Introduce standard work as compliance bureaucracy without involving operators in its definition. A standard not shared will not be respected.
Apply autonomous maintenance (TPM) as a prerequisite for advanced automation
Rationale: Automated lines require reliable equipment. Without autonomous maintenance (cleaning, inspection, first-level lubrication by operators), machines degrade rapidly and OEE collapses.
Do: Train operators on first-level inspections specific to their machines. Create daily visual inspection checklists (1-2 minutes per cycle). Integrate autonomous maintenance into standard work procedures.
Don't: Consider maintenance the exclusive domain of the maintenance department. The sharp divide between "who produces" and "who maintains" is one of the main obstacles to high OEE.
Use gemba walks as a structured managerial practice, not an occasional visit
Rationale: Management by walking around without structure does not produce actionable insights. A structured gemba walk with objectives, guiding questions, and systematic follow-up is a powerful improvement tool.
Do: Schedule regular gemba walks (daily for team leaders, weekly for managers). Focus on observing the process, not the people. Always bring paper and pen to note waste. Close the loop with documented and tracked actions.
Don't: Conduct gemba walks to "check" operators or verify procedure compliance. The goal is to observe the process, not to supervise people.
Implement SPC on critical characteristics before reducing inspection frequency
Rationale: Reducing inspections without in-process statistical monitoring increases the risk of undetected non-conformities. SPC ensures the process is in statistical control before reducing final inspections.
Do: Identify critical characteristics from PFMEA. Implement control charts (X-bar/R or I/MR depending on sampling approach). Achieve statistical control (absence of special causes) before lowering the sampling frequency of the final inspection.
Don't: Eliminate final inspections before the process demonstrates both capability (Cpk ≥ 1.33) and stability (control chart in statistical control) documented on at least 30 consecutive samples.
Size safety stock on real variability, not rules of thumb
Rationale: Excessive safety stock ties up capital; insufficient safety stock causes stockouts. The statistical formula based on actual demand and lead time variability is more accurate than any rule of thumb ("X weeks of coverage").
Do: Calculate safety stock as SS = z × σ_d × √LT, where z is the factor for the desired service level, σ_d is the standard deviation of demand, and LT is lead time. Revise periodically with updated data.
Don't: Apply uniform "2-week stock" policies to all items. Use ABC/XYZ classification to size differently items with different variability and importance.