GB560 Unit 9 Control Plan Example

First Control Limits From Twenty Pilot Days: A Unit 9 Control Plan for a Redesigned Returns Process

[Student Name]

Purdue University Global

GB560: Designing, Improving, and Implementing Processes

Unit 9 Control Plan

[Instructor Name]

[Date]

The seller, its returns center and every chart value are composite and illustrative; the methods and sources cited are real.

Chart 1: Daily Median Dock-to-Refund Time

Chart type. An individuals and moving range chart. The process produces one daily median, not subgroups of several measurements taken together, so an X-bar chart, which needs rational subgroups, does not fit.

Data. Twenty daily medians from the pilot, averaging 27.3 hours, with an average moving range of 2.95 hours.

Limits. Upper limit = 27.3 + 2.66 × 2.95 = 35.2 hours. Lower limit = 27.3 − 2.66 × 2.95 = 19.5 hours. The moving range chart's upper limit is 3.267 × 2.95 = 9.65 hours.

Target, kept separate. The Unit 2 requirement sets a three-business-day refund window that 95 percent of returns must meet. The control limits describe what the process currently does; the target describes what it must do. A day at 34 hours would be within limits and still well inside the target, while a sudden run of days at 20 hours would be a signal worth investigating even though it beats the target.

Control limits describe the process as it is; targets describe it as someone wants it to be, and confusing the two hides both problems and improvements.

Chart 2: Share of Refunds Corrected After Posting

Chart type. A p-chart, because each refund is either correct or corrected, and the number of refunds varies by day.

Center line. 0.41 percent of refunds corrected after posting during the pilot.

Limits at full volume. At about 290 refunds a day, the upper limit is 0.0041 + 3 × √(0.0041 × 0.9959 / 290), about 1.54 percent. At about 1,450 a week, the weekly upper limit is about 0.91 percent. The lower limit is zero in both cases.

Why both. Daily limits are wide because a single day has few corrections; the weekly chart can detect a smaller shift, so both are kept. Montgomery (2020) notes that a p-chart's sensitivity depends on sample size, which is why the weekly chart is the better guard on accuracy.

Measure 3: Exception-Lane Aging

Measure. The number of exception-lane items older than 48 hours, counted each morning.

Ceiling. Five. This is a fixed ceiling, not a statistical limit, because the measure is a count of items that should not exist; any item older than 48 hours means a shopper is waiting without a decision.

Response Table

Signal: one point beyond a control limit. Owner investigates the same day, records the cause and, if the cause is a special event such as a system outage, documents it; if not, escalates to the supervisor.

Signal: a run of eight consecutive days all above, or all below, the average. The process has likely shifted. The owner reviews what changed in the period, and the supervisor decides whether the shift is an improvement to keep or a problem to correct.

Signal: exception ceiling passed. The exception lead clears the oldest items first that morning and reports the cause to the supervisor by noon.

No signal. No action. Adjusting a stable process in response to ordinary variation makes it worse, a point Deming (1986) illustrated with his funnel experiment.

What the Pilot Data Show About Stability

Before limits are trusted, the data that produced them should come from a stable process. The twenty pilot medians were checked against the limits they produced: none fell outside, the longest run on one side of the average was five days, and no moving range exceeded 9.65 hours. The first two days of the pilot ran slightly high while staff learned the new sequence, but not beyond the limits, so they were kept. If they had been outside, the limits would have been recalculated without them and the reason recorded. The correction rate showed about five corrected refunds over the pilot's roughly 1,276 returns, too few to judge stability on their own, which is another reason the p-chart limits are set for full volume and will be revisited once full-volume data exist.

Owners and Review Rhythm

Dock-to-refund chart: the returns center supervisor, reviewed daily at the morning meeting.

Refund correction charts: the customer service lead, daily chart reviewed daily, weekly chart reviewed each Monday.

Exception aging: the exception lead, counted and reported every morning.

Monthly review: the operations director reviews all three with the supervisor.

Recalculating the Limits

The limits come from one product family during the pilot. They will be recalculated after twenty working days at full volume, once every family is on the new path, because the full process may behave differently. After that, limits change only when the process is deliberately changed or when enough new stable data has accumulated. Recalculating after every signal is a common error, since it absorbs the very shifts the chart exists to reveal. Wheeler and Chambers (1992) recommend holding limits fixed while the process is stable and revising them only with a documented reason.

References

Deming, W. E. (1986). Out of the crisis. MIT Press.

Montgomery, D. C. (2020). Introduction to statistical quality control (8th ed.). Wiley.

Wheeler, D. J., & Chambers, D. S. (1992). Understanding statistical process control (2nd ed.). SPC Press.