Choose Lean, Six Sigma, or a combined approach from the problem evidence.

Compare the primary questions each approach asks

Lean asks what the customer values, how work and information flow, where delay and waste occur, what prevents pull, and how people can improve the standard. Six Sigma asks how output varies, whether the process is capable, which inputs explain the variation, and how a verified control can hold the gain.

Neither definition should be made absolute. Lean uses data and statistical thinking; Six Sigma projects often improve flow. The useful distinction is the dominant uncertainty and the amount of analytical discipline required for the decision.

• Lean: value stream, lead time, queues, flow, pull, waste, standard work
• Six Sigma: defect definition, measurement system, variation, capability, causal factors, control
• Both: customer requirement, evidence, experimentation, ownership, and sustainment

Choose from the problem signature—not the preferred brand

Use lean methods when long lead time, batching, handoffs, excess WIP, imbalance, poor workplace organization, or unstable daily response dominates. Use Six Sigma when the output distribution, defect rate, process capability, measurement reliability, or multivariable relationship is central.

Use a combined approach when flow and variation interact. A process may have poor capability that creates rework queues; a scheduling change may expose product-dependent cycle variation. Begin with the least complex method capable of resolving the uncertainty, then add depth when evidence demands it.

Match the project path and data burden to risk

A lean PDCA cycle may use observation, time study, a current-state map, and a controlled trial. DMAIC adds formal definition, measurement-system analysis, baseline capability, statistical analysis, designed experiments where appropriate, and control planning. Do not impose every tool on every project.

Check data quality before analysis. A precise model built on mixed products, unstable definitions, biased sampling, or an inadequate measurement system produces false confidence. For low-risk reversible changes, rapid tests may teach more than months of analysis.

1. Define the gap

   State customer requirement, current result, scope, and consequence.

2. Inspect the signature

   Separate flow, stability, variation, measurement, and constraint effects.

3. Choose depth

   Use the simplest credible method for the risk and uncertainty.

4. Test causality

   Run a controlled change, statistical test, or designed experiment.

5. Control and learn

   Update flow rules, standards, controls, and response plans.

Integrate people and system context into the analysis

A statistically significant factor is not automatically the best operational intervention. Consider safety, ergonomics, maintenance, product mix, queues, buffers, the system constraint, and the people who must run the control. Simulation can test whether a local capability gain changes customer output.

Avoid certification hierarchy as project governance. The process owner, operators, subject experts, and decision owner need a shared view of evidence and risk. Specialists should strengthen the team’s learning, not remove ownership from the work.

Sustain the gain through visible operating control

For flow changes, define release rules, WIP limits, standard work, escalation, and customer measures. For variation changes, define measurement, control limits, reaction plans, maintenance, setup conditions, and product-specific boundaries.

Review the change across enough cycles, shifts, products, and disturbances. If the result depends on the project team watching constantly, the operating system has not absorbed it. Record what was predicted, what happened, and where the chosen method was insufficient.