Field Book · Chapter 8 of 8 · Capability, effectiveness and LPA

Prove It Held

One good shift is not proof. Set the criteria before you look, re-run capability, hold a 30 day window, and add a layered process audit question that keeps the fix in place.

Set the proof before you look: a fresh capability run of at least 20 subgroups, and no chart signals and no coolant readings over the limit for 30 days. Here Cpk comes back to 1.90 and the 30 days are clean before the corrective action closes. Then add one layered process audit question so the fix stays in place.

The lathe is running again and the chart looks fine. That is one good shift. Proof is a run long enough to be convincing, measured against criteria you set before you started.

Re-run capability

Fresh run, same lathe, coolant held at 21 to 22 degrees: 20 subgroups across four shifts, 100 pieces. Average 24.0039, average range 0.00655, sigma 0.00282. Cp is 2.36 and Cpk is 1.90.

Compare it with the stable stretch before the drift: Cp 2.03 and Cpk 1.63. The difference in Cp is within what samples of 55 and 100 pieces can show by chance, so call the spread unchanged. What matters is that Cpk is back, and well above the 0.98 of the drifted stretch.

Set the proof before you look

  1. A fresh capability run of at least 20 subgroups with Cpk of 1.67 or higher. Your customer may set another number.
  2. No mean beyond a limit and no run of eight for 30 days.
  3. The coolant log shows no reading above 25 and no unacknowledged chiller alarm for 30 days.

The first is met today. The other two need the calendar. At day 30 the chart has no point beyond a limit and no run of eight, the highest coolant reading in the log is 22 degrees, and the chiller alarm log is clean. Now the corrective action can close.

Add the question that keeps it fixed

One question joins the layered process audit for Lathe 3: "Is the coolant at the lathe at or below 25 degrees, and is there a reaction on the sheet for any reading above?" The shift lead asks it each shift, the supervisor weekly. Layers matter because the person who owns the process is the one least likely to see the slip.

Read more: how to run an effectiveness check and how to run a layered process audit.

The mistake to avoid

Closing the action on the day the fix goes in. A fix that has run one good shift has not been tested by a bad one. If it fails inside the window, go back to the cause, not to the next restart.

ON THE FLOOR MONDAY
  • Pick your last drift fix and find the proof it held.
  • If there is none, write the criteria and the window this week.
  • Add one question about it to your layered process audit.

Open the printable Monday Board

Want every tool on this trail in one purchase? The Process Trail Bundle has all 13.

The part, the machine, the shift and every number in this book (shaft SH-3306 on Lathe 3) are a made-up example. The tool sidebars show ShopFloorOps tools, which I make and sell. Every chapter teaches the step itself, so you can use your own forms. New to a term? See the quality glossary.