Control phase: standardization, monitoring and sustaining gains
Control makes the gain durable. It answers: who watches the process, with which metric, how often, on which chart, and what they do when the signal fires. The exam tests control chart choice, the difference between control and specification limits, and the practical artefacts of standardisation and handover.
The control plan
A control plan names each critical x and Y, the measurement method, sample size and frequency, the responsible person, the control method (chart, check sheet, poka-yoke) and the reaction plan. Without a reaction plan a chart is decoration.
Standardise and hand over
- Standard work: documented current best method, visible at the point of use, owned by the process owner.
- Training and competence checks for everyone affected; update job aids and work instructions.
- Visual management: andon, shadow boards and status boards that make abnormality obvious.
- Audit schedule: layered checks confirming the standard is followed - audit the process, not the people.
- Closure: benefits verified against the baseline, documentation filed, lessons learned, sponsor sign-off, ownership transferred.
Control charts
- Individual continuous measurements: I-MR chart.
- Continuous with rational subgroups (n = 2–10): Xbar-R, or Xbar-S for larger subgroups.
- Proportion defective: p chart (np for constant sample size).
- Defect counts: c or u chart. Control limits come from the process's own variation (±3 sigma), never from the customer specification. Points outside the limits, runs of seven or eight on one side, or clear trends signal special-cause variation: investigate, do not blindly adjust (tampering with common-cause variation increases variation).
Worked example
A support team answers 95% of chats within target. A manager proposes drawing limits at the 90% service-level agreement. Two errors: the SLA is a specification, not a control limit; and the daily proportion answered on time is attribute data suited to a p chart (a count of chats per agent would suit c or u). Correct setup: p chart of the daily on-time proportion, limits from 20–25 stable days, and a reaction plan naming who investigates what.
Traps that cost marks
- Placing specification limits on a control chart.
- Reacting to every up-and-down movement, which is tampering.
- Writing a control plan no one owns after the team disbands.
- Declaring savings that finance has not verified.