Improve and Control Phases: Solution Implementation and Sustainment
Domain 5 tests whether an improvement sticks. Questions cover solution selection, piloting, control plans, control-chart choice and the reaction rules that stop gains from decaying.
Improve
- Generate broadly, then screen with impact/effort, FMEA risk reduction and cost.
- Pilot before full rollout; use DOE - screening designs first - to find the X settings that move Y.
- Mistake-proofing hierarchy: eliminate the error, then prevent it, then detect it.
- Verify the gain statistically (repeat the hypothesis test) and financially, then update standard work and documentation.
Control
- Control plan: characteristic, specification, measurement method, sample size and frequency, owner, and the out-of-control action plan (OCAP).
- SPC limits come from the process (±3σ), never from specifications. Specification limits state what the customer wants; control limits state what the process actually does.
- Chart choice: X̄-R for subgrouped continuous data (n ≈ 2–5), I-MR for individuals, p for proportion defective, c or u for defect counts.
- Signal rules include a point beyond 3σ, 8–9 consecutive points on one side of the centre line, and 6 steadily trending points.
- Sustain with standard work, visual management, layered audits, training and a named process owner.
Worked example
A filling line runs subgroups of 5 every 30 minutes. X̄ = 250 mL and R̄ = 4 mL, so with A2 = 0.577 the X̄ limits are 250 ± 0.577 × 4, i.e. 247.7 to 252.3 mL. The last six subgroup means rise steadily. That is a trend signal even though every point lies inside the limits: investigate the filler head, do not widen the limits, and do not adjust a stable process point by point.
Traps that cost marks
Specification limits drawn on a control chart; tampering with common-cause variation; a chart that does not match the data type; no OCAP or named owner; declaring victory without post-improvement capability; letting the control plan lapse at handover.