# Six Sigma DMAIC: Control

1. Six Sigma DMAIC Control: Monitor the variation of the process after applying the improvement.
Control what: performance, cost, and risks.

Statistical Process Control: Statistical methods to control the process, using control charts.

Statistical Process Control-SPC: Any control system must have:
– A measurable goal/standard.
– A mean to measure improvement.
– The (potentially automatic) comparison between the goal and the current state to signal the need for corrective action.

Run chart: A plotting of measure against time: measure X time.
Control Chart: A Run Chart with Upper and Lower Control Limits (UCL and LCL). Invented by Shewhart to distinguish between different causes of variation:

Causes of Variation:

Common cause: stable/trend over time & predictable.
Special cause: unstable & unpredictable.

Process: In statistical control vs. out of control.

Control charts: X axe: time. Y axe: KPI. If the relationship between KPI and KPIV is proven, one can use the KPIV in the Y axe instead. Draw Upper Control Limit as horizontal lines. At least ±3 * Sigma upper & lower lines. Plots data points not averages.
UCL and LCL: Upper Control Limit vs Lower Control Limit.
UCL & LCL=X̅  ± 3S.
Method 1: Draw points, and study the pattern of these points –
Method 2: Draw bar charts, their center is the average of last 11 points, and their height is the sigma of the last 11 points.

Statistical Process Control: Steps:
1. Prepare: define the measure.
2. Control chart selection (graph)
3. Divide UCL & LCL Areas to: A B C.
4. Plot your data then check for one of the out-of-control rules:

Out-of-control rules:
1 case beyond LCL/UCL.
2 out of 3 consecutive cases in area A.
4 out of 5 consecutive cases in area B.
8+ consecutive cases in area C.
8+ consecutive cases with consistent deterioration/improvement.

Out-of-control patterns:

• Point outside control limit.
• Graduate or Sudden shift in process average.
• Cycles: disturbances with gradual causes with certain level of regularity.
• Trends: gradual deterioration or improvement in performance

SPC is only effective in early 6 sigma stages when variation is relatively large (not near 3.4 dpmo).

Statistical state of a process: Comparing process to control and tolerancelimits:

Ideal: process in statistical control and meeting customer tolerance (specifications) limits.

On threshold: meeting upper & lower control limits but not tolerance limits

On brink: meeting tolerance levels but not control limits.

Chaos: not meeting either control limits or tolerance limits.

Comparing Six Sigma to Quality Department data: Quality department data use child (statistic, average) distribution, made up of population of averages of multiple samples, so it is not KPIV distribution, thus its Sigma s=S/n (or S=s*n), and be careful: such smoothing can hide important changes in data.