Minitab Capability Analysis: Turning Measurement Data into Capability Decisions

Stop describing process performance with words like tight and loose

A shop floor might say a piston diameter is tight or a fill weight is loose, but those descriptions do not answer whether the process will stay inside the drawing’s specification limits. Engineers need a quantitative comparison between observed variation and the allowed tolerance. Minitab’s Capability Analysis provides that comparison by calculating indices such as Cp, Cpk, Pp, and Ppk, and by checking whether the data look normally distributed.

What the capability indices measure

Capability Analysis compares the width of the specification band to the spread of the measurement data.

• Cp – ratio of the tolerance width (USL minus LSL) to six times the process standard deviation. It tells you whether the spread fits inside the limits, ignoring where the process is centered.
• Cpk – the smaller of the upper-side and lower-side capability calculations. It penalizes a process whose mean is off-center.
• Pp – same tolerance-to-spread comparison, but based on overall variation including shifts between subgroups or over time.
• Ppk – centered version of Pp, using overall variation.

Values above 1.0 generally indicate the spread fits inside the tolerance band; values below 1.0 flag a potential capability problem. Because the indices are dimensionless, you can compare them across different dimensions and units.

Running Capability Analysis in Minitab

1. Open the Minitab worksheet containing the numeric measurement column, such as piston diameter in millimeters.
2. Choose Stat > Capability Analysis > Normal (or Cpk in older releases).
3. Select the measurement column and enter the Upper Specification Limit (USL) and Lower Specification Limit (LSL), for example USL = 100.0 and LSL = 95.0.
4. Click OK. Minitab displays Cp, Cpk, Pp, Ppk, and a normal probability plot.

You need permission to open and read the worksheet; the file must not be locked by another session. The analysis reads data and does not modify the worksheet, so no rollback is required.

Illustrative example: piston diameter

Suppose a technician collects 100 piston diameter measurements. The specification is USL = 100.0 mm and LSL = 95.0 mm, giving a tolerance width of 5.0 mm. After running Capability Analysis, Minitab might return:

Cp = 1.33
Cpk = 1.12
Pp = 1.30
Ppk = 1.10
Estimated sigma = 0.63 mm

Interpretation: the process spread is about 33 percent of the tolerance width, so the spread itself fits comfortably. However, Cpk is lower than Cp, which means the process mean is not centered between the specification limits. Moving the machine mean closer to 97.5 mm would raise Cpk toward the Cp value.

A manual check uses Cp = (USL - LSL) / (6σ). With sigma ≈ 0.63 mm, 6σ ≈ 3.8 mm, and Cp ≈ 5.0 / 3.8 ≈ 1.33, consistent with Minitab’s output. This example is illustrative; actual values depend on your data.

When the indices can mislead

• Normality assumption. Cp and Cpk assume the data follow a normal distribution. Skewed or heavy-tailed data can make the index look better than the process actually is.
• Sample size. Fewer than 30 observations produce unstable estimates. Treat results from small samples as preliminary.
• Subgrouping. If Ppk is much lower than Cpk, variation between subgroups or over time is hurting capability even if short-term spread looks good.

Always inspect the normal probability plot Minitab generates. Points that curve away from the straight line or fall far outside the band suggest non-normality or outliers. In those cases, consider a transformation, a different distribution model, or more data before making a change.

Practical verification checklist

• Confirm the measurement column is numeric and free of obvious entry errors.
• Check that the sample size is at least 30; 100 or more gives more stable indices.
• Review the normal probability plot for curvature, outliers, or stepped patterns.
• Manually compute Cp = (USL - LSL) / (6σ) and Cpk = min[(USL - μ)/(3σ), (μ - LSL)/(3σ)] and compare with Minitab’s output.
• If Cpk differs from Cp by more than about 0.1, investigate the mean location before acting on the index alone.

Actionable closing

Gather at least 30 measurements, run Capability Analysis in Minitab, record the indices and probability plot, and verify the output against a manual calculation. If Cpk is below your target, center the process first; if the plot shows non-normal behavior, fix the distribution assumption before trusting the number.