dmaic-control

Define each CTQ with its measurement method, frequency, named owner, UCL/LCL limits, and a specific response action for deviations. The Skill provides a template covering the Control Plan, SPC configuration, reaction plan, SOPs, training, and formal project closure.

Run the included check-control-limits.py script on a CSV file of sequential measurements. It computes mean, sigma, UCL, and LCL, then reports violations of all 8 Western Electric rules with 1-based point indices and reaction guidance.

For continuous data use I-MR (n=1), X-bar/R (n=2-9), or X-bar/S (n>=10). For attribute data use p-chart or np-chart for defective proportions and u-chart or c-chart for defect counts. The reference guide includes a full selection tree.

Control limits must be calculated from post-improvement data collected during the Improve pilot, never from the Measure baseline. You need at least 20-25 observations or subgroups from the stabilized improved process for reliable limits.

Applying all 8 rules simultaneously raises the false alarm rate to roughly 5% per point. The Skill recommends using Rules 1 and 4 as a minimum, adding more only for high-criticality processes, and calibrating based on process risk.