Process capability indices (Cp and Cpk) are statistical metrics that quantify how well a manufacturing process can consistently produce parts within specified tolerance limits. To break this down in plain language:
Cp and Cpk both compare the natural variation of a process to its tolerance limits, but they differ in one key way:
For example, a process with a tight spread but a mean far from the target will have a high Cp but a low Cpk.
Process capability directly impacts three core aspects of manufacturing: part function, assembly, and cost.
Critical dimensions (e.g., the diameter of a bearing bore, the depth of a thread) must stay within tolerance to ensure the part performs as designed. For example, a bearing bore that is too small will cause the bearing to bind, leading to premature failure; a bore that is too large will result in excessive play, causing vibration and noise. Cp and Cpk quantify the risk of producing parts outside tolerance: a Cpk of 1.33 corresponds to a defect rate of approximately 66 parts per million (PPM), while a Cpk of 1.67 corresponds to 0.57 PPM.
Consistent part dimensions simplify assembly and reduce rework. For example, if a bracket’s mounting hole diameter varies widely, assembly operators may struggle to align it with a mating part, requiring manual adjustment or even scrapping the part. This is particularly important for automated assembly lines, which cannot compensate for large dimensional variation.
Poor process capability increases costs in three ways:
For example, a CNC machine with high repeatability (a key driver of process capability) will produce more consistent parts. The Gree CNC GA-DV750 vertical machining center, with a repeatability of 0.0024 mm, is better suited for high-precision applications requiring tight tolerances than the GA-MV856, which has a repeatability of 0.006 mm.
Measuring Cp and Cpk requires collecting data on a process’s output, calculating the process’s spread and mean, and applying the appropriate formula. Achieving high process capability requires controlling the process’s variation and alignment.
To calculate Cp and Cpk, you need a sample of part dimensions produced under stable operating conditions. A stable process is one that is in statistical control, meaning its variation is consistent over time (no sudden shifts or trends).
Sample size is a critical factor in the accuracy of Cp and Cpk calculations. The standard recommendation for initial process capability studies is 30 samples, but this is a minimum. For processes with high variation, larger sample sizes (50–100 samples) are needed to get a reliable estimate of the process’s spread and mean.
For example, if you are evaluating the capability of the GA-UHD500 5-axis machining center (repeatability 0.004 mm) for a tight-tolerance application, collecting 50 samples will give you a more accurate measure of its capability than 30 samples.
Once you have a sample of data, calculate two key statistics:
The formulas for Cp and Cpk are:
To improve Cp and Cpk, you need to reduce process variation and align the process mean to the nominal dimension.
Variation in manufacturing processes comes from four main sources (the 4 Ms):
To align the process mean to the nominal dimension, adjust the process parameters (e.g., tool offset, machine position) to shift the mean of the part dimensions closer to the target. For example, if the mean diameter of a hole is 9.995 mm (nominal 10.000 mm), adjust the tool offset to increase the hole diameter by 0.005 mm.
Cp and Cpk calculations are only as accurate as the data and assumptions used to calculate them. Several sources of error can inflate or deflate Cp and Cpk values.
A small sample size (less than 30) can lead to an inaccurate estimate of the process’s spread and mean. For example, a sample of 10 parts may have a standard deviation that is much lower than the true process standard deviation, leading to an inflated Cp and Cpk.
Cp and Cpk assume that the process data is normally distributed (follows a bell curve). If the data is not normally distributed (e.g., skewed, bimodal), the Cp and Cpk values will be inaccurate. For example, a process that produces parts with a skewed distribution (more parts on one side of the mean) will have a Cpk that does not accurately reflect the true defect rate.
Cp and Cpk are only meaningful for processes that are in statistical control. If a process is unstable (e.g., has sudden shifts in the mean or spread), the Cp and Cpk values will not predict future process performance. For example, a process that has a sudden shift in the mean due to tool wear will have a Cpk that is no longer valid after the shift.
Measurement error (variation in the measurement system) can inflate the process standard deviation, leading to a deflated Cp and Cpk. For example, if the measurement system has a repeatability of 0.002 mm, and the process standard deviation is 0.003 mm, the total measured variation will be higher than the true process variation, leading to an inaccurate Cp and Cpk.
Cp and Cpk are useful metrics, but they have several limitations that are often overlooked.
Cp and Cpk measure dimensional variation, but they do not measure part function. A part may meet all dimensional tolerance requirements but still fail to function as designed. For example, a shaft may have a diameter within tolerance but still fail due to material fatigue.
Cp and Cpk are typically calculated from short-term data (e.g., a single shift). They do not account for long-term variation (e.g., tool wear, material changes, operator shifts) that can cause the process to drift over time. For example, a process that has a Cpk of 1.67 over a single shift may have a Cpk of 1.0 over a month due to tool wear.
Cp and Cpk are calculated for individual dimensions, but they do not account for interactions between dimensions. For example, two dimensions that are each within tolerance may interact to cause a part to fail. For example, the length and width of a bracket may each be within tolerance, but the angle between them may be out of tolerance, causing the bracket to not fit properly.
Cp and Cpk are often used as a sole measure of process performance, but they should be used in conjunction with other metrics (e.g., control charts, defect rates) to get a complete picture of process performance. For example, a process with a high Cpk may still produce defects if the tolerance limits are too loose.
| Method | Accuracy | Effort | Typical Use |
|---|---|---|---|
| Cp (Process Capability Potential) | High if process is centered and data is normal; low if process is shifted or data is non-normal | Low | Initial process capability assessment, comparing the potential of different machines (e.g., GA-DV750 vs GA-MV856) |
| Cpk (Process Capability Index) | High if process is stable, data is normal, and sample size is sufficient; low if process is unstable, data is non-normal, or sample size is small | Medium | Ongoing process monitoring, verifying process performance meets customer requirements |
| Control Charts (for process stability) | High | Medium | Verifying process is in statistical control before calculating Cp and Cpk |
| Gage R&R (for measurement error) | High | Medium | Assessing the accuracy of the measurement system used to collect data for Cp and Cpk |
What is a good Cp and Cpk value? A Cp and Cpk of 1.33 is generally considered the minimum acceptable value for most manufacturing processes. A Cp and Cpk of 1.67 is considered excellent and is often required for safety-critical applications.
Can Cpk be higher than Cp? No, Cpk can never be higher than Cp. Cp measures the best-case capability of the process (perfectly centered), while Cpk measures the actual capability of the process. If the process is perfectly centered, Cpk will equal Cp. If the process is shifted, Cpk will be lower than Cp.
How does sample size affect Cp and Cpk? A small sample size can lead to an inaccurate estimate of the process’s spread and mean, leading to inflated or deflated Cp and Cpk values. A larger sample size (50–100 samples) will give a more accurate estimate of the process’s capability.
What is the difference between Cp/Cpk and Pp/Ppk? Cp and Cpk are short-term capability indices that are calculated from data collected over a short period (e.g., a single shift) and assume the process is in statistical control. Pp and Ppk are long-term capability indices that are calculated from data collected over a longer period (e.g., a month) and do not assume the process is in statistical control. Pp and Ppk are often used to assess the overall capability of a process over time.
Can I use Cp and Cpk for non-normal data? Cp and Cpk are designed for normally distributed data. If the data is not normally distributed, the Cp and Cpk values will be inaccurate. For non-normal data, you can use non-normal capability indices (e.g., Pp, Ppk) or transform the data to make it normal before calculating Cp and Cpk.
In conclusion, Cp and Cpk are valuable tools for assessing process capability, but they must be used correctly and with an understanding of their limitations. By following the practical checklist and considering the sources of error and limitations outlined in this article, you can use Cp and Cpk to improve process performance and reduce costs.
Send us your part, batch size and target port. We reply with a configuration, a binding lead window and a total-cost comparison.
Request a quote →