Calculate process capability indices (Cp, Cpk, Pp, Ppk) with DPMO, sigma level, and distribution visualization
Process capability analysis is one of the most powerful tools in quality engineering and Six Sigma methodologies. At its core, it answers a deceptively simple question: is your manufacturing process actually capable of producing parts that meet your customer's specifications? The answer lies in a family of dimensionless indices — most notably Cpk, Cp, Ppk, and Pp — that quantify the relationship between how spread out your process is and how wide the allowed tolerance band is. The Cpk (Process Capability Index) is the most widely used single-number summary of process performance. A Cpk of 1.33 means your process produces parts within spec at a level corresponding to roughly 4-sigma quality, with only about 63 defects per million opportunities (DPMO). A Cpk of 2.0 represents world-class six-sigma quality with fewer than 0.002 DPMO. Most automotive and aerospace industries require Cpk ≥ 1.67 for critical dimensions, while general manufacturing typically sets the bar at Cpk ≥ 1.33. Understanding the difference between the four main indices is crucial. Cp and Cpk use within-subgroup (short-term) standard deviation, reflecting the process's inherent variation when operated consistently. Pp and Ppk use overall standard deviation, capturing all sources of variation including shifts, drifts, and special causes over time. The 'k' in Cpk and Ppk stands for centering — these indices account for how well the process mean is centered within the specification window. If your Cp is high but Cpk is significantly lower, it means your process is precise enough but is not properly centered; the fix is simply to adjust the process mean rather than reduce variation. This calculator supports two input modes. In Summary Statistics mode, you enter the process mean, standard deviation (within or overall), and specification limits directly. This is ideal when you already have statistical summaries from an SPC system or control chart software. In Raw Data mode, you paste your actual measurement values (comma or newline separated) and the calculator computes mean and standard deviation automatically. Raw data mode also calculates both within-group and overall standard deviations when you specify a subgroup size, allowing simultaneous computation of all four indices. Beyond the core indices, this calculator provides a complete quality picture: sigma level (the process's effective sigma quality level), DPMO (estimated defects per million opportunities derived from the normal distribution), and process yield (the percentage of conforming output). The financial impact simulator lets you translate these abstract statistical measures into concrete business terms by estimating annual non-conforming units and the cost of poor quality. The bell curve visualization overlays your process distribution against the specification limits, making it immediately clear how much of the distribution falls outside the tolerance band. The capability bar chart compares Cp, Cpk, Pp, and Ppk side by side against the industry-standard thresholds of 1.00 (minimum acceptable) and 1.33 (capable), giving process engineers a quick visual read on improvement priorities. Whether you are a quality engineer setting up a new production line, a manufacturing manager reviewing supplier capability data, a Six Sigma Black Belt conducting a process improvement DMAIC project, or a student learning statistical process control, this tool provides the comprehensive capability analysis you need — entirely in your browser with no data leaving your device.
Understanding Process Capability
What Is Process Capability?
Process capability measures how well a production process performs relative to its specification limits. A capable process consistently produces output within the customer-defined tolerance band. The capability indices Cp and Cpk quantify this using the within-subgroup (short-term) standard deviation, while Pp and Ppk use the overall standard deviation to capture long-term process performance including all sources of variation. Cpk and Ppk are the 'centered' versions of Cp and Pp respectively — they penalize processes whose mean is not centered within the specification window. The specification limits — Upper Specification Limit (USL) and Lower Specification Limit (LSL) — define the acceptable range; any output outside these limits is a nonconformance or defect.
How Are Capability Indices Calculated?
The core formula is elegantly simple: Cp = (USL − LSL) / (6σ), where σ is the within-subgroup standard deviation. This compares the specification width to six times the process spread (which covers ±3σ, or 99.73% of a normal distribution). Cpk adds centering: Cpk = min[(USL − mean) / (3σ), (mean − LSL) / (3σ)]. The two components are CPU (upper capability) and CPL (lower capability); Cpk takes the minimum, so the worse side of the distribution determines the index. Pp and Ppk use the same formulas but substitute the overall sample standard deviation (s = √[Σ(xi − x̄)² / (n−1)]) for σ. The Taguchi Cpm index extends this further: Cpm = Cp / √(1 + [(mean − target)² / σ²]), penalizing processes that are off-target even if within spec.
Why Does Capability Analysis Matter?
Capability analysis translates statistical process behavior into actionable quality decisions. A Cpk below 1.0 means the process is actively producing defects and requires immediate corrective action. A Cpk between 1.0 and 1.33 is marginally capable — defects are occurring at a low rate but improvement is recommended. A Cpk of 1.33 or higher is the general industry standard for acceptable production quality, corresponding to about 63 DPMO at 4-sigma. Automotive suppliers under AIAG standards must demonstrate Cpk ≥ 1.67 for critical characteristics before production approval. Beyond compliance, capability analysis guides investment decisions: if Cp >> Cpk, the process has sufficient precision but needs centering — a cheap adjustment. If Cp itself is too low, the fundamental process variation must be reduced — typically a more expensive engineering intervention.
Limitations et mises en garde
Capability indices assume the process follows a normal (Gaussian) distribution. For non-normal data — common in tool wear, surface finish, or biological measurements — standard Cpk calculations can be misleading, and transformation or non-parametric methods are needed. A minimum of 30 data points is recommended for statistical validity; with fewer points, the standard deviation estimate is unreliable and Cpk confidence intervals are wide. Capability indices also assume the process is in a state of statistical control (stable over time). If your process has trends, shifts, or out-of-control signals on a control chart, fix those issues before interpreting Cpk. Finally, never confuse specification limits with control limits: control limits are statistical signals derived from process data, while specification limits are customer requirements — they are completely independent concepts.
Formules
Compares the specification width to the process spread (6σ). Cp ≥ 1.33 means the spec window is at least 33% wider than the process spread. Does not account for centering.
Takes the minimum of upper capability (CPU) and lower capability (CPL), penalizing an off-center process. Cpk ≤ Cp always; they are equal only when the process is perfectly centered.
Defects Per Million Opportunities estimated from the normal distribution. Φ is the standard normal CDF. A Cpk of 1.33 yields ~63 DPMO; a Cpk of 2.00 yields ~0.002 DPMO.
Extends Cp by penalizing deviation from a target value T. When X̄ = T, Cpm = Cp. As the mean drifts from target, Cpm decreases even if all output is within specification.
Reference Tables
Cpk Thresholds and Corresponding Quality Levels
| Cpk | Sigma Level | DPMO | Yield (%) | Industry Interpretation |
|---|---|---|---|---|
| 0.33 | 1σ | 317,310 | 68.27 | Extremely poor — immediate action required |
| 0.67 | 2σ | 45,500 | 95.45 | Poor — high defect rate |
| 1.00 | 3σ | 2,700 | 99.73 | Minimum acceptable — process just fits spec |
| 1.33 | 4σ | 63 | 99.9937 | Capable — general manufacturing standard |
| 1.67 | 5σ | 0.57 | 99.99994 | Highly capable — automotive critical features |
| 2.00 | 6σ | 0.002 | 99.9999998 | World class — Six Sigma target |
Industry Cpk Requirements
| Industry / Standard | Minimum Cpk | Contexte |
|---|---|---|
| General manufacturing | 1.33 | Standard production quality level |
| Automotive (AIAG PPAP) | 1.67 | Critical and significant characteristics |
| New process launch | 1.67 | Run-at-rate qualification criterion |
| Aerospace (AS9100) | 1.33–1.67 | Depending on criticality classification |
| Medical devices (ISO 13485) | 1.33 | For validated production processes |
| Six Sigma programs | 2.00 | World-class quality target |
Worked Examples
Shaft Diameter — Centered Process
Specification width = USL − LSL = 25.10 − 24.90 = 0.20 mm
Cp = 0.20 / (6 × 0.015) = 0.20 / 0.09 = 2.22
CPU = (25.10 − 25.00) / (3 × 0.015) = 0.10 / 0.045 = 2.22
CPL = (25.00 − 24.90) / (3 × 0.015) = 0.10 / 0.045 = 2.22
Cpk = min(2.22, 2.22) = 2.22 (perfectly centered: Cp = Cpk)
DPMO ≈ 2 × Φ(−6.67) × 10⁶ ≈ 0.0 (effectively zero defects)
Fill Volume — Off-Center Process
Specification width = 505 − 495 = 10 mL
Cp = 10 / (6 × 1.5) = 10 / 9 = 1.11
CPU = (505 − 503) / (3 × 1.5) = 2 / 4.5 = 0.44
CPL = (503 − 495) / (3 × 1.5) = 8 / 4.5 = 1.78
Cpk = min(0.44, 1.78) = 0.44
The process is shifted high — mean is 3 mL above nominal (500 mL)
Financial Impact of Improving Cpk
At Cpk = 1.00: DPMO = 2,700 → defects/year = 500,000 × 2,700/1,000,000 = 1,350
Annual cost at Cpk = 1.00: 1,350 × $25 = $33,750
At Cpk = 1.33: DPMO = 63 → defects/year = 500,000 × 63/1,000,000 = 31.5 ≈ 32
Annual cost at Cpk = 1.33: 32 × $25 = $800
Annual savings = $33,750 − $800 = $32,950
How to Use the Cpk Calculator
Choisissez votre mode d'entrée
Select 'Summary Statistics' if you already know your process mean and standard deviation (e.g., from SPC software or a control chart). Select 'Raw Data' to paste your actual measurement values — the calculator will automatically compute mean, standard deviation, and sample size for you.
Enter Specification Limits and Process Data
Type your Upper Specification Limit (USL) and Lower Specification Limit (LSL) from your engineering drawing or customer requirement. Then enter the process mean and standard deviation (Summary mode) or paste your comma/newline-separated measurements (Raw Data mode). Optionally enter a Target/Nominal value to compute the Cpm Taguchi index.
Review Capability Indices and Charts
The calculator instantly displays Cp, Cpk, Pp, Ppk, CPU, CPL, sigma level, DPMO, and process yield. The ProgressRing shows your Cpk on a scale of 0–2 with color coding (red < 1.0, yellow 1.0–1.33, green ≥ 1.33). The capability bar chart compares all four indices against the 1.00 and 1.33 threshold lines.
Interpret Results and Estimate Financial Impact
Read the capability assessment text for your process status and recommended action. If you have annual production volume and cost-per-defect data, enter those in the Financial Impact section to see projected annual non-conforming units and cost of poor quality — useful for building a business case for process improvement investment.
Questions Fréquemment Posées
What is the difference between Cpk and Ppk?
Cpk uses the within-subgroup (short-term) standard deviation, estimated from rational subgroups on a control chart. It reflects the process's inherent, best-case capability when operating in statistical control. Ppk uses the overall sample standard deviation calculated from all measurements — it captures all sources of variation including shifts, drifts, and special causes that occur over time. Cpk is a snapshot of what the process can achieve; Ppk is what the process actually achieves over the long run. Typically Ppk ≤ Cpk. A large gap between them signals the process is not stable and may have assignable causes affecting performance. Both indices account for centering; their non-centered counterparts are Cp and Pp respectively.
What Cpk value is considered acceptable?
The most widely accepted threshold is Cpk ≥ 1.33 for general manufacturing, which corresponds to approximately 4-sigma quality and 63 defects per million opportunities (DPMO). Automotive suppliers under AIAG APQP standards must demonstrate Cpk ≥ 1.67 for critical and significant characteristics before production approval (PPAP). New process launches often require Cpk ≥ 1.67 as a run-at-rate qualification criterion. The Six Sigma benchmark is Cpk ≥ 2.00, representing fewer than 0.002 DPMO. At the minimum, Cpk ≥ 1.00 means the process spread just fits within the specification window when perfectly centered, but any shift in the mean will produce defects — this is the absolute floor for 'marginally acceptable.'
My Cp is high but my Cpk is low — what does that mean?
This is actually good news from a troubleshooting perspective. A high Cp with a lower Cpk means your process has sufficient precision — the natural spread of your process (6σ) is narrower than your specification width — but the process mean is not centered within the tolerance band. In other words, you are making consistent parts, just not centered on target. The fix is typically straightforward: adjust the process mean (shift a tool offset, change a setpoint, recalibrate a fixture) to center the distribution between LSL and USL. This is almost always cheaper than reducing variation. The Cpk improvement from centering is immediate and requires no capital investment.
How many data points do I need for a reliable Cpk calculation?
The general recommendation is a minimum of 30 measurements for a reliable Cpk estimate. With fewer points, the sample standard deviation has high uncertainty — the 95% confidence interval on Cpk with only 10 data points is extremely wide. The AIAG Measurement System Analysis (MSA) manual and PPAP requirements typically require 25–30 subgroups for control chart-based capability studies. For preliminary process capability studies, 50–100 measurements provide a robust estimate. This calculator will display a warning when raw data mode is used with fewer than 30 measurements, alerting you that the results are indicative rather than statistically reliable. Always collect data under representative production conditions.
What is DPMO and how is it calculated from Cpk?
DPMO stands for Defects Per Million Opportunities — the expected number of defective units if you produced one million parts under the same process conditions. It is calculated from the normal distribution using the Cpk value: for a two-sided specification, DPMO = [Φ(−3×CPU) + Φ(−3×CPL)] × 1,000,000, where Φ is the standard normal cumulative distribution function. Equivalently, DPMO ≈ 2 × Φ(−3 × Cpk) × 1,000,000 as an approximation. The corresponding yield is 1 − DPMO/1,000,000. A Cpk of 1.33 gives roughly 63 DPMO and 99.9937% yield. A Cpk of 2.00 (six sigma) gives approximately 0.002 DPMO. Note that the traditional 3.4 DPMO for six sigma includes a 1.5σ long-term shift allowance.
What is the Cpm (Taguchi) index and when should I use it?
The Cpm index (also called the Taguchi capability index) extends Cpk by penalizing processes that are off-target — even if the mean is still within specification. The formula is Cpm = Cp / √(1 + [(mean − target)² / σ²]). When the process mean equals the target, Cpm equals Cp. As the mean drifts away from target (even while staying within spec), Cpm decreases. Cpm is most useful when the target value is not the midpoint of the specification (asymmetric tolerances) or when quality loss increases continuously as the characteristic deviates from the nominal — the Taguchi quality loss function philosophy. For symmetric tolerances where being anywhere within spec is acceptable, standard Cpk is sufficient.