Process Excellence

Find the cause of an operating problem and design an improvement that works for the whole process. Use Lean, Six Sigma, or a simpler diagnostic according to the question and evidence. A narrow analysis does not require producing every DMAIC artifact.

Establish boundaries, customer requirement, decision, available data, and the user's requested output. Work from the supplied baseline and definitions. If data is incomplete, provide a provisional diagnosis or measurement plan with limitations; do not invent a baseline or require real-world sign-offs before drafting. Execution gates and controlled rollout still matter where the operational risk requires them.

Measure the work as it happens

Define the start and end event, unit of work, eligible cases, period, and whether time includes waiting, nonbusiness hours, rework, and incomplete cases. Separate touch time from elapsed time. Inspect missing events, duplicates, time zones, selection bias, and open cases before trusting averages.

Map the flow, decisions, queues, handoffs, and exceptions from evidence. SIPOC can establish boundaries; value-stream mapping can expose waiting and inventory. Use the mapping depth that helps the decision. Do not infer a bottleneck solely from the longest average duration when parallel capacity, batch size, or demand differs.

Use measurement and statistical methods for event logs, capability, control charts, and flow measures. Apply statistical tools only when their assumptions and data are suitable. Unknown performance is not a sigma level.

Test causes and design improvements

Use observation, process data, interviews, Pareto analysis, fishbone, and repeated why questions to generate testable explanations. Distinguish a suspected cause from a verified cause. “Human error” may hide a design or control issue; an external cause should not be discarded merely because the team cannot control it.

Assess waiting, rework, unnecessary movement, excess work, inventory, redundant processing, and unused capabilities in context. Do not remove a control or apparently non-value-adding step without understanding its purpose and obligation. Check downstream effects and constraints before optimizing one step.