First steps
- Check stability and normality.
- If Cp is fine and Cpk is not: re-centre.
- If Cp is low: reduce variation (multi-vari, DOE) or review the tolerance with the customer.
How to measure success
Cp, Cpk and Ppk on 125 or more values. Record the value before you start, so the effect of every action can be shown.
When these methods are the wrong choice
- Process capability (Cpk): Unstable processes; make them stable first.
- MSA / Gage R&R: Nothing: do it before you trust any data.
- Design of experiments (DOE): Problems whose cause is already known, or processes you cannot run for trials.
- SPC control charts: Characteristics without a capable gauge or with too few data.
- Multi-vari study: Problems already narrowed to two or three factors.
- Hypothesis tests: Exploring causes; use it to confirm them.
The methods in one line
| Method | What it is |
|---|---|
| Process capability (Cpk) | Compares the process spread and centring with the tolerance and predicts PPM. |
| MSA / Gage R&R | Checks that the measurement system can see the variation you want to control. |
| Design of experiments (DOE) | Tests several process settings at once to find the factors and interactions that drive the result. |
| SPC control charts | Separates common-cause from special-cause variation and tells the operator when to react. |
| Multi-vari study | Finds whether variation lives within the part, part to part or over time before you start experiments. |
| Hypothesis tests | Proves whether a difference (before / after, machine A / B) is real or chance. |