Four key figures, two pairs of letters, constant confusion. Initial sample inspection reports list Ppk values, production reports list Cpk values, and machine acceptance reports list Cmk—yet no one explains why a particular metric is required in each context. This leads to follow-up questions, discussions with the customer, and, in the worst case, rejected sample submissions—even though the process was actually in order.
The difference between Cp, Cpk, Pp, and Ppk is smaller than it seems once you understand the two underlying questions: Which type of variation is being measured—short-term or long-term—and is the location of the mean taken into account or not? These two questions form a simple framework into which all four metrics can be clearly classified. Once you’ve internalized this framework, you’ll be able to read any test report without having to look anything up.
This article clarifies both questions and shows when each metric is required. You’ll learn how short-term and long-term variation differ, what the gap between the metrics reveals about your process, which metric is required in which production phase, and according to which thresholds the evaluation is performed. This will enable you to confidently select the correct metric in the future and explain to the customer why.
KEY POINTS AT A GLANCE
|
IN SHORTThe four metrics differ along two axes: type of variation and reference to the mean. Cp and Cpk use the short-term variation from subgroups, while Pp and Ppk use the overall variation. Cp and Pp assume a centered process, while Cpk and Ppk take the actual location into account. Cpk is the standard for production monitoring, while Ppk is used for initial sampling. |
|
CONTENTS OF THIS ARTICLE |
All four are capability indices. They express the width of a process relative to the allowed tolerance. The difference lies in two characteristics: which measure of dispersion is used and whether the mean is taken into account.
| Metric | Variation | Mean | Time Horizon |
|---|---|---|---|
| Cp | Short-term (within subgroups) | ignored | short |
| Cpk | Short-term (within subgroups) | taken into account | short |
| Pp | Long-term (overall dispersion) | ignored | Long |
| Ppk | Long-term (total dispersion) | taken into account | long |
Short-term variation is estimated from the variation within the subgroups, usually over the mean range of the control chart. Long-term variation is the classic standard deviation across all individual values. Therefore, long-term variation is almost always greater, and Pp and Ppk tend to be smaller than Cp and Cpk.
The confusion often stems from the fact that the four metrics appear in different combinations in different documents. The initial sample inspection report shows Ppk, the production reporting shows Cpk, and the machine acceptance report shows Cmk. Anyone familiar with the underlying system can interpret each of these documents correctly and immediately recognize which type of variation is being referred to.
A vivid image helps with understanding: You can think of the four metrics as two photos of the same process—one taken with a short exposure time and one with a long exposure time. The “short-exposure” image, based on Cp and Cpk, shows the process in sharp focus at a single moment, while the “long-exposure” image, based on Pp and Ppk, shows it blurred over time, capturing all its movements. Both images of the same process answer different questions and complement each other.
Cp and Cpk describe how capable a process is in a steady state. They use the internal variation within the subgroups—that is, the variation that occurs from part to part under otherwise identical conditions.
Cp is the ratio of the tolerance width to six times the short-term variation. Cpk is the smaller of the two one-sided values and takes into account how far the mean has shifted from the center.
In production monitoring, Cpk is the standard metric because the process is already established at that stage and short-term variation is the relevant factor. As long as the control chart shows stability, Cpk describes the capability that can be expected over the long term. If stability is lost, the Cpk also loses its significance, and the focus shifts to the Ppk.
A practical note on the data basis: Cp and Cpk are only as reliable as the control chart from which their short-term variation is derived. If the variation is estimated from too few or irregularly recorded subgroups, both metrics will fluctuate from analysis to analysis. A stable control chart maintained over time is therefore a prerequisite for robust short-term metrics.
Pp and Ppk describe a process’s actual performance over an extended period, including all fluctuations between shifts, batches, and tool changes. They utilize the total variation of all individual values.
Because long-term variation includes short-term variation as well as drift over time, Ppk is generally smaller than Cpk. A large gap between the two indicates a process that, while currently running within tight tolerances, drifts over time. It is precisely this drift that the initial sampling is intended to reveal, which is why the PPAP process requires Ppk.
The value of Ppk lies precisely in its ability to portray reality unvarnished. While Cpk answers the question of how well the process performs in the best-case scenario, Ppk answers the question of what actually comes off the line over the course of weeks. For the customer, the second question is the more important one, because they receive the actual shipment, not the ideal process.
This has a clear implication for the initial sampling. Because there is still little historical data available, Ppk is the more honest metric, as it does not assume a level of stability that has not yet been proven. Only when the process has been proven to run stably over several weeks does Ppk approach Cpk, and the analysis may then shift to the short-term metric.
Cp and Cpk indicate what a process is capable of when everything is running smoothly. Pp and Ppk indicate what it actually delivers over the course of weeks. The difference is the cost of instability.
Amadeus Lederle, Chief Technology Executive, CSP Intelligence GmbH
| Observation | Meaning | Recommended Action |
|---|---|---|
| Cp close to Cpk | Process is well-centered | Maintain centering |
| Cp significantly greater than Cpk | Mean value is off-center | Adjust the process to center |
| Cpk close to Ppk | Process stable over time | Continue monitoring |
| Cpk significantly greater than Ppk | Process drifts over time | Eliminate the causes of the drift |
Together, these four metrics form a diagnostic tool. The difference between Cp and Cpk measures off-center variation, while the difference between Cpk and Ppk measures drift over time. By placing all four side by side, you can quickly identify—in a matter of seconds—where a process’s problem lies: in the variation, in the location, or in the stability over time. No single value can do this on its own.
The choice of metric depends on the phase. During pre-production and initial sampling, there is still limited data available, and the customer wants to see actual performance, including all variations. That is why Ppk is required at this stage. In ongoing production, the process is established and stable, and monitoring is based on Cpk.
| Phase | Required metric | Rationale | Source |
|---|---|---|---|
| Initial Sample Approval (PPAP) | Ppk | Actual performance, including all variations | AIAG-VDA PPAP |
| Production Monitoring | Cpk | Short-term capability of the stable process | AIAG-VDA SPC |
| Machine Acceptance | Cmk | Capability of the machine in isolation | VDA Volume 5 |
| Process Validation | Cpk and Ppk | Both horizons in the documentation | IATF 16949 8.5.1 |
A practical rule of thumb helps: short-term for monitoring, long-term for demonstration. In ongoing monitoring, the focus is on whether the process is currently under control; the short-term analysis using Cpk is suitable for this. When demonstrating performance to the customer, the focus is on actual performance over time; this is where Ppk comes into play. Both are part of a complete process validation.
In process validation, both time horizons are intentionally required because together they provide a complete picture. Cpk demonstrates that the process is capable when in a controlled state, while Ppk demonstrates that it remains so over the actual operating period. Only together do they bridge the gap between laboratory conditions and everyday production and provide the customer with the assurance needed for approval.
The evaluation thresholds are identical for all four k-metrics because they all refer to the same ratio of process width to tolerance.
| Value | Evaluation | Range |
|---|---|---|
| Less than 1.00 | Not capable | Immediate action required |
| 1.00 to 1.32 | Conditionally capable | Needs improvement |
| 1.33 | Capable | Automotive Industry Standard |
| 1.67 | Proficient | Critical characteristics |
PRACTICAL NOTENever confuse the required metric. A customer who requires a Ppk of 1.67 will not accept a Cpk of 1.67, because the Cpk ignores time drift. Check the specifications carefully to determine exactly which metric is required and at what threshold. |
It is noteworthy that the evaluation thresholds apply regardless of the type of metric. This is no coincidence but follows from the definition: All four relate the same process width to the same tolerance, but with different sources of variation. Therefore, a Ppk of 1.33 implies the same defect rate as a Cpk of 1.33, relative to the respective time horizon.
Manually maintaining four key performance indicators across many characteristics and production lines is prone to errors. The CSP Manufacturing OS automatically calculates Cp, Cpk, Pp, and Ppk from the recorded process data. The software clearly distinguishes between short-term and long-term variation and documents each metric in an audit-proof manner for audit purposes.
The key advantage of a shared database is consistent calculation. When Cpk and Ppk are derived from the same raw data and automatically calculated separately for short-term and long-term variation, the typical discrepancies that arise from manual maintenance are eliminated. The customer then sees consistent figures that they can trust without question.
Cpk uses the short-term variation within the subgroups, while Ppk uses the total variation over the entire period. Ppk therefore also includes drift over time and is smaller than Cpk in unstable processes.
Cp considers only the variation and assumes a perfectly centered process. Cpk additionally takes the mean into account. Therefore, Cpk is never greater than Cp.
In the initial sample inspection according to the AIAG-VDA PPAP process, Ppk is usually required because it reflects the actual process performance, including all variations. In series production, monitoring is then performed using Cpk.
Ppk uses the total variation, which includes not only internal variation but also the drift of the process over time. This additional variation increases the denominator and reduces the value.
Yes. Both are evaluated using the same thresholds: 1.33 or higher indicates capability, and 1.67 or higher indicates secure control. The only difference is which type of variation is included in the calculation.
Cmk is the machine capability index. It isolates the influence of the machine under constant conditions and is used during machine acceptance testing in accordance with VDA Volume 5. The requirement here is usually 1.67.
No. All four require a normal distribution. For other distributions, distribution-specific methods according to ISO 22514 are used.