Process Capability & Cpk Calculator

Compare two processes against shared specifications using the NIST sample-estimator formulas. Enter ordered observations or a sample mean and sample standard deviation for each process. Review the limiting tolerance margin, centers and spreads, observed outside counts and the mean or spread that would meet your chosen index goal. Lower or upper specifications can be evaluated separately.

Process worksheet

Compare process capability against your specifications

Compare two samples or summaries, inspect which specification is limiting, and see what centering or spread would reach your chosen goal.

Only Process B meets the entered index goal.

Process A Cpk
1.1094
The upper specification is limiting.Observed sample · 51 readings
Process B Cpk
1.3333
Equal margins to both limits.Entered summary · 60 readings
Shared specifications

Both processes must use this unit. The selector labels values without conversion.

Process A

Paste 2–250 values in order, separated by commas, tabs or new lines.

Process B

Describe the summary sample. The indices use the entered mean and spread.

The NIST sample index interpretation assumes stable, approximately normal processes. This worksheet compares estimates and keeps observed counts separate from future expectations.

Process centers and spread
  • Lower spec 45 mm
  • Upper spec 55 mm
  • Target 50 mm
Process means and three-sample-standard-deviation bandsShared measurement scale. Process A mean 52 mm, sample standard deviation 0.901388 mm. Process B mean 50 mm, sample standard deviation 1.25 mm.ABMeasurement (mm)

Each dot is a mean, and each bar spans mean ± 3s. The brown dashed lines mark specification limits. The slate dotted line marks the target when entered.

Center or narrow the process

Keep one estimate fixed and see what value of the other would reach Cpk 1.25.

Process A

Maximum spread at current mean
0.8 mm
Mean at current spread
48.380204 to 51.619796 mm.

Process B

Maximum spread at current mean
1.333333 mm
Mean at current spread
49.6875 to 50.3125 mm.
Readings in entered order
  • Lower spec 45 mm
  • Upper spec 55 mm
  • Target 50 mm
Observed sample run orderProcess A has 51 supplied readings, 0 below, 51 inside and 0 above specification.Reading orderValue (mm)

Squares mark observations outside an active specification. These specification lines are not control limits.

Process A
0 below · 51 inside · 0 above
Open capability ledger
Capability and sample summaries by process
MeasurementProcess AProcess B
Sample basisObserved sample · 51 readingsEntered summary · 60 readings
Mean52 mm50 mm
Sample standard deviation0.901388 mm1.25 mm
Lower specification margin7 mm5 mm
Upper specification margin3 mm5 mm
Cpk1.10941.3333
Cp1.8491.3333
Cpl2.58861.3333
Cpu1.10941.3333
Cpm0.75971.3333
Observed inside51Unavailable
Observed outside0Unavailable
Open observations ledger
Observed readings by process and entry order
ProcessOrderReadingStatus
Process A150.5 mminside
Process A251.375 mminside
Process A352.25 mminside
Process A453.125 mminside
Process A550.875 mminside
Process A651.75 mminside
Process A752.625 mminside
Process A853.5 mminside
Process A951.25 mminside
Process A1052.125 mminside
Process A1153 mminside
Process A1250.75 mminside
Process A1351.625 mminside
Process A1452.5 mminside
Process A1553.375 mminside
Process A1651.125 mminside
Process A1752 mminside
Process A1852.875 mminside
Process A1950.625 mminside
Process A2051.5 mminside
Process A2152.375 mminside
Process A2253.25 mminside
Process A2351 mminside
Process A2451.875 mminside
Process A2552.75 mminside
Process A2650.5 mminside
Process A2751.375 mminside
Process A2852.25 mminside
Process A2953.125 mminside
Process A3050.875 mminside
Process A3151.75 mminside
Process A3252.625 mminside
Process A3353.5 mminside
Process A3451.25 mminside
Process A3552.125 mminside
Process A3653 mminside
Process A3750.75 mminside
Process A3851.625 mminside
Process A3952.5 mminside
Process A4053.375 mminside
Process A4151.125 mminside
Process A4252 mminside
Process A4352.875 mminside
Process A4450.625 mminside
Process A4551.5 mminside
Process A4652.375 mminside
Process A4753.25 mminside
Process A4851 mminside
Process A4951.875 mminside
Process A5052.75 mminside
Process A5152 mminside
Download process capability CSV
Show selectable CSV

How to use this calculator

  1. Choose two-sided, lower-only or upper-only specifications. Enter the active limits, an engineering target for two-sided checks and a relevant index goal.
  2. Use one measurement unit. Enter each process as ordered readings or as a sample mean and standard deviation.
  3. Compare the limiting index and signed specification margins. Inspect the sampled mean and spread on the common scale.
  4. Read the goal panel to see the theoretical spread at the current mean or allowable centers at the current spread.
  5. Inspect the entered-order plot and observation ledger where raw data are provided. Download calculations and source observations as CSV.

The Formula

The mathematical logic powering this calculation is:

Use the NIST sample estimates. For observed values, the mean x̄ is their arithmetic average and s is the sample standard deviation with denominator n−1. Summary inputs use the entered x̄ and s directly. Two-sided indices are Cp = (USL − LSL)/(6s), Cpl = (x̄ − LSL)/(3s), Cpu = (USL − x̄)/(3s), Cpk = min(Cpl, Cpu), and Cpm = (USL − LSL)/(6√(s² + (x̄ − target)²)). A lower-only specification uses Cpl. An upper-only specification uses Cpu. A mean outside its limiting specification can give a negative index. At zero spread the worksheet withholds index and goal estimates.

For goal G, keeping the mean fixed allows a maximum positive s of the nearest specification margin divided by 3G. If the margin is zero or negative no positive s can meet a positive goal at that mean. Keeping s fixed requires x̄ ≥ LSL + 3Gs, x̄ ≤ USL − 3Gs, or both, depending on the active limits. If those conditions give an empty interval, centering alone cannot meet the goal with this spread. The user supplies G.

Observed values equal to an active limit count within specification. The entered sample is the only basis for observed counts. The mean ±3s chart summarizes sample spread. Neither those bands nor specification lines are statistical control limits. Index interpretations assume stable processes and approximately normal data with enough independent measurements, often at least 50. This worksheet does not test stability, fit a distribution, estimate defect rates, or compute a pooled within-subgroup standard deviation. Some systems call all-reading estimates Pp and Ppk. This worksheet labels the NIST sample formulas and states the input basis.

Use one unit for all measurements. The unit selector labels values without conversion. Measurements and bounds accept −1,000,000 to 1,000,000 with up to six decimals. s is nonnegative. Enter 2–250 observed values per process. An optional summary count is a whole number from 2 to 1,000,000. The positive index goal ranges from 0.01 to 10 with up to three decimals.

Worked Example

If you input:

  • Lower, upper and target: 45 mm, 55 mm, and 50 mm
  • Cpk goal: 1.25
  • Process A: 51 illustrative readings with mean 52 mm and sample s about 0.901388 mm
  • Process B: Summary mean 50 mm, sample s 1.25 mm from 60 values

Result:

Process A has Cpk 1.1094, constrained by the upper spec despite its smaller spread. At its current mean the most spread allowed for the goal is 0.8 mm. At its current spread the mean must be 48.380204–51.619796 mm. Process B has Cpk 1.3333 and meets the entered 1.25 goal.

Frequently Asked Questions

How can Cp be higher than Cpk?

Cp measures the full specification width relative to the spread. Cpk uses the smaller margin from the mean. The two are equal for a centered mean.

What does a negative capability index indicate?

A negative applicable index means the mean is already outside its limiting specification. Reducing spread around that same mean does not reach a positive index goal.

What happens when all measurements are equal?

The sample s is zero. The worksheet shows the observed counts, mean and zero spread and leaves indices and goals undefined.

Why are no defect rates predicted?

Observed counts are available for pasted readings. Extrapolating them requires more information about process stability, measurement quality and the distribution of future results.

What is the difference between an engineering target and midpoint?

A midpoint is halfway between specification limits. The target is the intended nominal value you enter. Cpm uses distance from this target, even when it differs from the midpoint.

Which standard deviation belongs in the summary?

For comparison with the readings mode use the sample standard deviation over the observations. A pooled within-subgroup estimate answers a different variation question. Choose a consistent basis across the two entries.

Similar Calculators