Cp / CpkProcess Capability Index
The fundamental statistical indices for measuring whether a manufacturing process is capable of meeting its quality specifications.

On this page · 8 sections
- 1Introduction to Process Capability
- 2Understanding Cp: Potential Process Capability
- 3Understanding Cpk: Actual Process Performance
- 4Acceptance Thresholds: The Meaning of 1.33 and 1.67
- 5Cp/Cpk vs. Pp/Ppk: Capability vs. Performance
- 6The Dangers of Misapplication: Normality and Stability
- 7Industry Applications: PPAP and Validation
- 8Advanced Pitfalls and Limitations
How does Cp / Cpk apply to your shop floor?
Pick your industry and scale — Ask V5 rewrites the definition in your context, gives a worked example, and shows what V5 does on day one.
01Introduction to Process Capability
In the world of regulated manufacturing—encompassing pharmaceuticals, medical devices, and high-precision electronics—producing quality goods is not merely about staying 'within the lines' at the moment of inspection. True quality excellence requires a deep understanding of process capability: the inherent ability of a process to produce output that conforms to specifications over time. Process capability analysis provides the numerical evidence needed to prove that a manufacturing system is robust, predictable, and reliable.
At the heart of this analysis are two primary indices: Cp and Cpk. These statistics allow quality engineers to translate complex variability into simple, actionable ratios. By comparing the 'voice of the process' (the natural spread of the data) to the 'voice of the customer' (the specification limits defined by the design team), these indices offer a snapshot of risk. A process with a high capability index has a significant safety buffer, meaning it is unlikely to produce non-conforming parts even if minor fluctuations occur. A process with a low index, however, is a liability that requires constant, expensive oversight.
Modern regulatory frameworks, such as the FDA's Quality System Regulation (21 CFR 820) and the ICH guidelines for pharmaceuticals, increasingly demand that manufacturers move beyond simple pass/fail testing. Instead, they must demonstrate 'Process Performance' through quantitative metrics. This shift ensures that quality is built into the process rather than just inspected out of the final product.
02Understanding Cp: Potential Process Capability
Cp is known as the 'potential' process capability index. It represents the best possible performance a process could achieve if its mean were perfectly centered between the Upper Specification Limit (USL) and the Lower Specification Limit (LSL). It focuses exclusively on the spread or 'width' of the process variation.
The mathematical definition of Cp, as provided by the NIST Engineering Statistics Handbook, is the ratio of the tolerance to the natural process spread:
"Cp = (USL - LSL) / 6σ"
In this equation, 'USL - LSL' represents the allowable tolerance range. The denominator, '6σ' (six sigma), represents the natural variation of the process. In a normal distribution, approximately 99.73% of all measurements fall within ±3 standard deviations from the mean. Therefore, a 6σ spread captures nearly all the output of a stable process.
- Cp > 1.0: The process spread is narrower than the specification limits. The process is potentially capable.
- Cp = 1.0: The process spread exactly matches the specification limits. There is zero margin for error.
- Cp < 1.0: The process spread is wider than the specification limits. The process will inevitably produce defects, regardless of how well it is centered.
The primary weakness of Cp is its 'blindness' to the process location. A process might have a very small standard deviation (high Cp) but still be producing 100% defective items if the entire distribution is shifted outside the limits. For this reason, Cp is used during initial design phases to determine if a chosen machine or method is *capable* of meeting the tolerance, but it is insufficient for monitoring actual production.
03Understanding Cpk: Actual Process Performance
Cpk is the index that accounts for the process mean. It measures the distance between the process center and the closest specification limit. This makes it a more realistic measure of 'actual' performance, as it penalizes the index if the process drifts toward a failure point.
NIST defines Cpk as the minimum of two one-sided capability indices:
"Cpk = min [ (USL - μ) / 3σ, (μ - LSL) / 3σ ]"
In this formula, μ (mu) is the process mean. By evaluating both (USL - μ) and (μ - LSL), Cpk identifies the side of the distribution most at risk of failing. The division by 3σ (rather than 6σ) reflects that we are looking at only one half of the distribution at a time.
The relationship between Cp and Cpk is a vital diagnostic tool for quality engineers:
- If Cpk = Cp: The process is perfectly centered on the target.
- If Cpk < Cp: The process is off-center. The gap between the two values indicates how much capability is being 'lost' due to poor centering.
- If Cpk = 1.0: The tail of the distribution is just touching a specification limit (roughly 0.135% defects on that side).
- If Cpk < 0: The process mean has drifted outside the specification limits, meaning more than 50% of the product is likely non-conforming.
In practice, Cpk is the 'gold standard' metric reported in management reviews and supplier quality reports. It tells the story of whether the process is *actually* succeeding in the present moment.
04Acceptance Thresholds: The Meaning of 1.33 and 1.67
What constitutes a 'good' Cpk? While a value of 1.0 mathematically suggests that the process is 'capable,' industry standards demand higher values to provide a safety margin against the inevitable 'shifts and drifts' of manufacturing.
| Cpk Value | Sigma Level | Expected Yield | Defect Rate (PPM) |
|---|---|---|---|
| 0.67 | 2σ | 95.45% | 45,500 PPM |
| 1.00 | 3σ | 99.73% | 2,700 PPM |
| 1.33 | 4σ | 99.9937% | 63 PPM |
| 1.67 | 5σ | 99.99994% | 0.57 PPM |
| 2.00 | 6σ | 99.9999998% | 0.002 PPM |
Industry-standard benchmarks have emerged over decades of statistical study:
- Cpk = 1.33: This is the typical requirement for established, stable processes. It corresponds to a 4-sigma level. It provides enough room that if the process mean shifts slightly, the parts will still be within spec.
- Cpk = 1.67: Often required for new processes during validation (e.g., PPAP) or for critical-to-safety features in medical devices. This 5-sigma level accounts for the fact that a new process may not yet have shown its full range of variability.
- Cpk = 2.00: The goal of Six Sigma programs. At this level, the process is so robust that even significant environmental changes are unlikely to cause a defect.
05Cp/Cpk vs. Pp/Ppk: Capability vs. Performance
One of the most frequent sources of confusion in quality engineering is the difference between 'Process Capability' (Cp/Cpk) and 'Process Performance' (Pp/Ppk). The difference lies in how variation—the standard deviation (σ)—is calculated.
Short-Term Capability (Cp/Cpk)
These indices use a 'short-term' estimate of standard deviation, often calculated from the average range (R-bar) or the pooled standard deviation within subgroups. This method effectively 'removes' the variation that occurs between different batches or time periods. It answers the question: 'How well does the process perform over a brief, stable window?' It represents the *potential* of the system under ideal control.
Long-Term Performance (Pp/Ppk)
Pp and Ppk use the 'overall' standard deviation (s), calculated from every single data point in the study. This captures all sources of variation: changes in humidity, different operators, different raw material lots, and mechanical wear. Ppk is what the customer actually experiences. It is generally more conservative (lower) than Cpk.
Comparing the two is a powerful diagnostic. If your Cpk is 1.8 but your Ppk is 1.2, you have a 'stability' problem. Your process is capable of excellence, but you are failing to keep it centered over the long term. This usually points to a need for better training, better environmental controls, or more frequent calibration.
06The Dangers of Misapplication: Normality and Stability
A Cpk value is only as good as the data and assumptions behind it. Applying these formulas blindly to any set of numbers can lead to dangerously wrong conclusions. There are two non-negotiable prerequisites for capability analysis:
1. The Process Must Be Stable
If a process is not in a state of statistical control—meaning it shows 'special cause' variation like sudden spikes or steady trends—the Cpk value is meaningless. A process that is jumping around today cannot be predicted for tomorrow. You must use control charts (like X-bar and R charts) to prove stability *before* calculating capability indices.
2. The Data Must Be Normally Distributed
The '6σ' and '3σ' denominators in the formulas are derived from the geometry of the normal (bell-shaped) distribution. If your data is skewed (for example, impurity levels that can't go below zero but have a long tail toward the high side), these formulas will give you a Cpk that significantly overestimates or underestimates your actual risk.
NIST warns that 'Large enough [sample size] is generally thought to be about 50 independent data values' to get a reliable estimate of capability. 1
07Industry Applications: PPAP and Validation
Capability indices are not just theoretical exercises; they are contractually and regulatorily mandated in many sectors.
- Production Part Approval Process (PPAP): In the automotive supply chain, a Ppk of 1.67 is often required for the 'initial process study' to gain approval for mass production. Once the process is running, a Cpk of 1.33 must be maintained.
- Medical Device Validation (PQ): Under 21 CFR 820.75, processes that cannot be 100% inspected (like sterile barrier sealing or plastic injection molding) must be validated. A high Cpk is the standard 'acceptance criterion' for a successful Process Qualification (PQ).
- Pharmaceutical CPV: The FDA's 2011 Process Validation guidance introduced 'Continued Process Verification' (Stage 3). Pharma manufacturers are now expected to monitor process trends, including Cpk, throughout the entire commercial lifecycle of a drug to detect potential quality shifts before they lead to a batch failure.
For a medical device company, a Cpk study might involve taking samples from the beginning, middle, and end of a production run. If the Cpk is high enough, the company can justify 'reduced inspection' or 'skip-lot testing,' significantly lowering operational costs while maintaining high confidence in safety.
08Advanced Pitfalls and Limitations
Even for experienced quality professionals, Cp and Cpk have nuances that can lead to errors in judgment.
One-Sided Specifications
If a requirement only has a maximum (e.g., 'Weight must not exceed 500mg'), you cannot calculate a Cp because there is no 'width' between limits. In these cases, you only calculate CPU (Capability Upper). Cpk is then simply equal to CPU.
The 'Target' Problem
Cpk rewards a process for being away from the limits, but it doesn't necessarily reward it for being exactly on the target value. This led to the creation of Cpm (the Taguchi capability index), which includes a 'penalty' for being off-target in the denominator. If your process needs to be precise (e.g., the fit of a piston), Cpm may be a better metric than Cpk.
Confidence Intervals
Cpk is an estimate based on a sample. A Cpk of 1.33 calculated from 30 parts has a wide margin of error. A true statistical report should include the 95% confidence interval (e.g., 'Cpk is 1.33 with a 95% CI of 1.15 to 1.51'). If the lower bound of the confidence interval is below your requirement (e.g., 1.0), you cannot truly claim your process is capable.
Finally, beware of the '1.5-sigma shift' convention often cited in Six Sigma literature. This convention assumes that over the long term, a process mean will drift by 1.5 standard deviations. This means a '6-sigma' process (Cp=2.0) is essentially expected to perform like a '4.5-sigma' process (Cpk=1.5) over years of operation. Whether this specific shift applies to your process is a matter of debate, but the principle of long-term drift is universally true.
Frequently asked questions
Q.Can a process be 'in control' but not 'capable'?+
Yes. Statistical control only means the process is consistent and predictable. If that consistent output is still outside the specification limits (because the machine isn't precise enough or the limits are too tight), the process is in control but incapable.
Q.Why do we divide by 6 sigma in the Cp formula?+
In a normal distribution, the range of ±3 standard deviations (a total of 6 sigma) covers 99.73% of the data. This is traditionally considered the 'natural spread' of a process. By dividing the tolerance by this spread, we see how many times the process can fit into the allowable limits.
Q.What is a 'negative' Cpk?+
A negative Cpk occurs when the process mean is actually outside of the specification limits. This indicates that more than 50% of the production is non-conforming. It is a sign of a critical process failure.
Q.What should I do if my data is not normally distributed?+
First, investigate if the non-normality is due to a mixture of two processes or an outlier. If the distribution is naturally non-normal (like a Weibull or Lognormal distribution), you must use a data transformation (like Box-Cox) or use a non-parametric method for calculating capability, such as the ISO 22514-2 percentile method.
Q.Is Cpk enough for a quality release decision?+
Cpk is a powerful indicator, but it should not be the only factor. A release decision should also consider the results of attribute testing, visual inspections, and a review of the batch's audit trail to ensure no deviations occurred that aren't captured by the variable data.
Primary sources
Further reading
- Statistical Process Control (SPC)Understand the control charts that must precede any capability study.
- Pp and Ppk Performance IndicesLearn about the long-term counterparts to Cp and Cpk.
- Measurement Systems Analysis (MSA)If your gauge is not capable, your Cpk values are unreliable. Learn why.
- Continued Process Verification (CPV)How the FDA expects manufacturers to monitor capability throughout the product lifecycle.
Explore this topic
Cp / Cpk sits inside 2 overlapping topic clusters in our glossary. Every neighbour is one click away.
Root-cause toolkit, SPC, capability and the rest of the QA practitioner's bench.
Want to see how Cp / Cpk could fit into your own records and workflows? Explore the related V5 pages or talk to our team about what applies to your operation.
