A capability index maps to a predicted fraction of parts outside the limits, assuming a normal, centered process and counting both tails. Use it to translate an index into a number your team can picture.
about 66.1 PPM (6.61e-3%)
Guide
What Cpk actually tells you
Cpk answers one question: how many process spreads fit between the mean and the nearest specification limit. A Cpk of 1.0 means the nearest limit sits exactly three standard deviations from the mean, so roughly 0.13% of parts fall outside on that side. A Cpk of 1.33 buys four standard deviations of margin, which is why it became the default customer minimum in automotive and general manufacturing.
The index only means something when the process behind it is stable. If your measurements drift shift to shift, the single number flatters you. That is exactly what the Cpk to Ppk comparison is for: when the two diverge, look at stability before celebrating the Cpk.
Cp vs Cpk vs Pp vs Ppk
Index
Variation used
Centering
Question it answers
Cp
Within-subgroup
Ignores centering
Could the process fit if perfectly centered?
Cpk
Within-subgroup
Accounts for centering
Does the process fit where it actually sits?
Pp
Overall
Ignores centering
What is the long-term potential, drift and all?
Ppk
Overall
Accounts for centering
What does the customer actually receive over time?
The formulas
Cp
(USL − LSL) / (6 × σwithin)
Cpk
min(USL − mean, mean − LSL) / (3 × σwithin)
Pp
(USL − LSL) / (6 × σoverall)
Ppk
min(USL − mean, mean − LSL) / (3 × σoverall)
σwithin is the pooled subgroup standard deviation, corrected with the c4 unbiasing constant (the default capability method used by Minitab). σoverall is the plain sample standard deviation of all the values, which is why Pp and Ppk capture drift between subgroups that Cp and Cpk deliberately ignore. In both, mean is the average of your measurements.
How many samples do you need
Capability indices are estimates, and small samples make wide confidence intervals. With 30 values, a computed Cpk of 1.42 could plausibly sit anywhere from about 1.1 to 1.7. Treat anything under 25 values as a first look, aim for 50 or more before quoting a number to a customer, and use 100 or more readings in rational subgroups for a PPAP-grade study. More data does not fix an unstable process; it just measures it more honestly.
Frequently asked questions
What is a good Cpk value?
A Cpk of 1.33 is the common minimum in automotive and general manufacturing, giving four standard deviations of margin to the nearest limit. 1.67 is typical for new launches and safety-critical characteristics. Below 1.0 the process is expected to produce out-of-specification parts.
What is the difference between Cpk and Ppk?
Cpk uses within-subgroup variation, so it describes the potential of a stable process. Ppk uses overall variation across the whole dataset, so it captures drift between subgroups. When Cpk and Ppk diverge, the process is probably not stable and you should look at that before trusting the Cpk.
What PPM does a Cpk of 1.33 give?
Assuming a normal, centered process and counting both tails, a Cpk of 1.33 corresponds to roughly 66 parts per million outside the specification. A Cpk of 1.0 is about 2,700 PPM and 1.67 is under 1 PPM. Use the converter on this page to check any value.
Do I need subgroups to calculate Cpk?
This calculator does. Cp and Cpk use within-subgroup (short-term) variation. With rational subgroups that comes from the pooled subgroup spread; on individuals data it is conventionally estimated from the average moving range instead, an I-MR style study this page does not run. So if you paste a single column with no subgroup size, this calculator reports Pp and Ppk only and tells you why, rather than mislabelling overall variation as Cpk.
How many measurements do I need for a capability study?
Treat fewer than 25 values as a first look. Aim for 50 or more before quoting a number to a customer, and use 100 or more readings in rational subgroups for a PPAP-grade study. Small samples give wide confidence intervals, so a single computed index can be misleading.
Is my data uploaded anywhere?
No. The entire calculation runs locally in your browser tab. Nothing you paste is sent to a server, so you can use it on controlled or customer data without a signup.
