Engineering calculators

Cpk Calculator

Updated Aug 17, 2026 By Jehan Wadia
Rate Formulas
Measurement Units
Data values, mean, σ, USL and LSL must all be entered in this unit.
Results and charts are converted into this unit. Capability indices are dimensionless.
Separate values with commas, spaces, tabs or new lines — paste straight from Excel or a CSV. Non-numeric entries are skipped and reported.
Valid values: 0

Specification Type
Between 3 and 50 bins. The histogram redraws immediately.
Must be greater than zero.
Specification Type
Capability you want to reach (must be greater than zero).
Specification Type
Symmetric allowance either side of the mean.

Run Chart — Check Process Stability First
Distribution Histogram with Fitted Normal Curve
Card A — Descriptive Statistics
Mean (μ)
Standard Deviation (σ)
Sample Size (n)
Specification Width (USL − LSL)
Card B — Capability Indices
Cpk
Cp (potential)
CPU
CPL
Ppk (simplified)
Uses overall sample σ (n−1); true Cpk requires within-subgroup σ
Process Sigma Level (Z)
Z = 3 × primary index
DPMO
Defects per million opportunities
Estimated Non-Conforming Rate
Centering / Limit Distance
Process Distribution vs. Specification Limits
Why Cpk Differs From Cp — Your Process vs. Perfectly Centred
Step-by-Step Solution

Introduction

This Cpk calculator shows how well your process fits inside your spec limits. Cpk is a process capability index. It compares your process spread and center to the Upper Spec Limit (USL) and Lower Spec Limit (LSL). A higher Cpk means fewer bad parts.

You can use it three ways:

  • Raw Data: Paste your measurements from Excel or a CSV. The tool finds the mean and standard deviation for you.
  • Mean & Std. Dev.: Already have your stats? Just type them in.
  • Target Planner: Pick a Cpk goal, like 1.33, and see the standard deviation your process needs to hit it.

You get Cp, Cpk, CPU, CPL, Ppk, sigma level, DPMO, and the percent of parts outside spec. Charts show a run chart, a histogram with a fitted normal curve, and a bell curve against your limits. A step-by-step section shows every formula, so you can check the math or use it in a report.

It works with two-sided, upper-only, or lower-only specs, and with units like mm, inches, mils, or no units at all. Use it for machining, molding, assembly, or any job where parts must stay in tolerance. For related shop-floor metrics, see the OEE calculator, the true position calculator, and the TRIR calculator.

How to use our Cpk Calculator

Enter your process data or your mean and standard deviation, plus your spec limits. The Cpk calculator returns your Cpk, Cp, CPU, CPL, sigma level, DPMO, scrap rate, step-by-step math, and charts of your process capability.

Input Unit: Pick the unit you typed your numbers in. All values you enter — data, mean, sigma, USL and LSL — must use this one unit. If your prints mix systems, the mm to inches calculator and the unit converter calculator can convert them first.

Result Display Unit: Pick the unit you want results and charts shown in. The tool converts for you. Cpk and Cp have no units, so they never change.

Tab choice: Use Raw Data if you have measurements. Use Mean & Std. Dev. if you already know your stats. Use Target Planner to find the spread you need to hit a Cpk goal.

Enter Data Values (Raw Data tab): Paste or type your measured parts. Split values with commas, spaces, tabs, or new lines. You can paste straight from Excel. You need at least 2 numbers. Click Load Sample Data to try it out. To review the same data set other ways, try the statistics calculator, the mean median mode calculator, or the outlier calculator.

Specification Type: Choose Two-Sided when the part has both a high and low limit. Choose Upper-Only or Lower-Only when just one side matters, like flatness or strength.

Upper Specification Limit (USL): Type the largest value a part can have and still pass.

Lower Specification Limit (LSL): Type the smallest value a part can have and still pass. It must be less than the USL.

Histogram Bin Count: Set how many bars the histogram uses, from 3 to 50. Use the plus and minus buttons. More bars show more detail; fewer bars show the shape better with small samples.

Process Mean (μ): Type the average of your process. This tells the tool where your parts sit between the limits. You can find it fast with the average calculator.

Process Standard Deviation (σ): Type your process spread. It must be greater than zero. The variance calculator is handy if you only have variance on hand.

Target Cpk (Target Planner tab): Type the Cpk you want to reach, such as 1.33 or 1.67. The tool shows the largest sigma that still meets it.

Two-Sided Input Basis (Target Planner tab): Choose Tolerance half-width to enter a plus/minus value from the mean, or Specification limits to enter the USL and LSL yourself.

Tolerance Half-Width (±): Type how far a part can drift each way from the mean and still pass.

Calculate and Reset: Results update as you type, but you can click Calculate anytime. Click Reset to load the default example again.

What Is Cpk?

Cpk is a process capability index. It tells you how well a process makes parts that fit inside your spec limits. It compares the spread of your process to the room you have between the Lower Spec Limit (LSL) and the Upper Spec Limit (USL). It also looks at where your average sits between those limits.

A high Cpk means most parts land safely inside spec. A low Cpk means parts are drifting out and scrap or rework is coming.

The Cpk Formula

Cpk uses two half-checks and keeps the worse one:

  • CPU = (USL − mean) ÷ (3 × σ) — room on the top side
  • CPL = (mean − LSL) ÷ (3 × σ) — room on the bottom side
  • Cpk = the smaller of CPU and CPL

Here σ (sigma) is the standard deviation, which measures how much your parts vary.

Cpk vs. Cp

Cp only asks: is my process spread narrow enough for the tolerance? It ignores where the average sits. Cpk asks the harder question: is my spread narrow enough and is it aimed at the middle?

If Cp is high but Cpk is low, your spread is fine but your process is off-center. Shifting the mean back toward the middle can fix that fast. If both are low, you need to cut variation.

What Cpk Values Mean

CpkMeaningRoughly
Below 1.00Not capableLots of defects
1.00 – 1.32MarginalTight, risky
1.33 – 1.66CapableCommon goal in industry
1.67 and upHighly capableVery few defects

Many auto and aerospace customers ask for Cpk ≥ 1.33. Some safety-critical parts ask for 1.67 or 2.00.

Sigma Level and DPMO

Sigma level is just Z = 3 × Cpk. A Cpk of 1.33 equals a 4-sigma process. DPMO means "defects per million opportunities." It turns your capability number into an easy count: how many bad parts you would expect out of a million. The tail math behind DPMO is the same idea used by the Z score calculator and the normal distribution calculator.

Cpk vs. Ppk

Cpk uses within-subgroup sigma, which shows short-term spread. Ppk uses overall sample sigma, which shows long-term spread across shifts, lots, and tool changes. When you only have one column of data and no subgroups, the number you get is really Ppk. Ppk is usually a bit lower than Cpk, since it includes drift over time.

Before You Trust the Number

Cpk only works if a few things are true:

  • The process is stable. Check a run chart first. If there are trends, jumps, or wild swings, fix those before judging capability.
  • The data is close to normal. A histogram with a fitted bell curve shows this. Skewed data (like flatness or roundness) needs a different method.
  • You have enough parts. Aim for 30 or more measurements. Fewer than that gives a shaky answer — the sample size calculator and the confidence interval calculator help you judge how firm your estimate is.
  • Your gauge is good. If the measuring tool has too much error, it inflates sigma and drags Cpk down for no real reason. A percent error calculator is a quick way to size up gauge bias against a master part.

One-Sided Specs

Some features only have one limit. Surface roughness has a max. Weld strength has a min. In those cases there is no Cp, and you report only CPU (upper-only) or CPL (lower-only). Use the matching setting so the math does not invent a second limit.

How to Raise a Low Cpk

  • Center the process. Adjust the tool offset or setpoint so the mean sits near the middle of the tolerance. This is often the quickest win.
  • Cut variation. Control temperature, material lots, fixtures, and operator method. Lower sigma raises both Cp and Cpk. On cutting jobs, steady feeds and speeds and consistent surface footage go a long way.
  • Fix the gauge. Run a Gage R&R study and remove measurement noise.
  • Open the tolerance. Only if engineering confirms the wider spec still works for the part.

Quick Example

Say a shaft spec is 49.6 mm to 50.4 mm. Your mean is 50.05 mm and sigma is 0.105 mm. CPU = (50.4 − 50.05) ÷ (3 × 0.105) = 1.11. CPL = (50.05 − 49.6) ÷ (3 × 0.105) = 1.43. Cpk = 1.11, the smaller one. Cp would be 1.27. The gap between 1.27 and 1.11 is caused by the process running high. Aim it 0.05 mm lower and Cpk jumps to about 1.27 with no other change.


Formulas used

Sample Mean
\bar{x} = \frac{\sum_{i=1}^{n} x_i}{n}
Sample Standard Deviation
s = \sqrt{\frac{\sum_{i=1}^{n}(x_i - \bar{x})^2}{n-1}}
Potential Capability (Cp)
C_p = \frac{USL - LSL}{6\sigma}
Upper and Lower Capability Indices
CPU = \frac{USL - \mu}{3\sigma}, \qquad CPL = \frac{\mu - LSL}{3\sigma}
Process Capability Index (Cpk)
C_{pk} = \min\left(\frac{USL - \mu}{3\sigma},\ \frac{\mu - LSL}{3\sigma}\right)
Process Sigma Level
Z = 3 \times C_{pk}
Non-Conforming Fraction and DPMO
p_{total} = P\!\left(Z > \frac{USL-\mu}{\sigma}\right) + P\!\left(Z > \frac{\mu-LSL}{\sigma}\right), \qquad DPMO = p_{total} \times 10^{6}
Required Standard Deviation for a Target Cpk
\sigma_{req} = \frac{d}{3 \times C_{pk,target}}, \quad d = \min(USL-\mu,\ \mu-LSL)

Frequently asked questions

Can Cpk be a negative number?

Yes. Cpk goes negative when your average sits outside a spec limit. For example, if the USL is 50.4 and your mean is 50.6, the top side is already broken.

A negative Cpk means more than half your parts fail. Fix the process center right away.

Can Cpk ever be larger than Cp?

No. Cpk is always equal to or lower than Cp.

They match only when the mean sits exactly in the middle of the tolerance. Any shift off center pulls Cpk down while Cp stays the same.

What does Cpk = 0 mean?

It means your average is sitting right on a spec limit. About 50% of parts would fall outside on that side.

Stop and re-center the process before making more parts.

Does this calculator use n or n−1 for standard deviation?

It uses n−1, the sample standard deviation. That is the standard choice for capability studies with a sample of parts.

Using n instead would make sigma look smaller and Cpk look better than it really is.

Why is my Cpk different from the number in Minitab or my SPC software?

Most SPC software splits data into subgroups and uses within-subgroup sigma for Cpk. This tool uses overall sigma from one column of numbers.

That makes the result here closer to Ppk. It is usually a little lower, which is the safer number to report.

Why doesn't my DPMO match the Six Sigma table I have?

Many Six Sigma tables add a 1.5 sigma shift for long-term drift. That is why they show 3.4 DPMO at 6 sigma.

This calculator uses the plain normal curve with no shift, so it reports the short-term number. Both are correct; they just answer different questions.

How many parts should I measure?

Aim for at least 30 measurements. Fifty to 100 is better.

The tool will run with as few as 2 values, but small samples give a shaky Cpk that can swing a lot from one batch to the next.

Why do I get an error when all my values are the same?

If every number is identical, the standard deviation is zero. Cpk needs to divide by sigma, and you cannot divide by zero.

This usually means your gauge is too coarse. Use a gauge with more decimal places so it can see real part-to-part change.

Can I paste data straight from Excel or Google Sheets?

Yes. Copy a column or row and paste it into the Raw Data box.

Commas, spaces, tabs, and new lines all work. Text like part numbers or blank cells gets skipped, and the tool tells you how many it skipped.

Do all my numbers need to be in the same unit?

Yes. Your data, mean, sigma, USL, and LSL must all use the Input Unit you picked.

Mixing mm and inches in one entry gives a wrong answer. Convert first, then enter.

Does changing the display unit change my Cpk?

No. Cp, Cpk, CPU, and CPL are ratios, so the units cancel out.

Only the mean, sigma, spec width, and chart labels change when you switch the display unit.

How many histogram bins should I use?

A good starting point is the square root of your sample size. For 35 parts, try about 6 bins. For 100 parts, try about 10.

Too many bins makes the shape look choppy. Too few hides skew and double peaks.

What is the run chart for?

It plots your parts in the order you made them. It answers one question: is the process stable?

If you see a trend, a step jump, or wild swings, the Cpk number is not trustworthy yet. Fix the instability first, then judge capability.

My calculator says 0.5% out of spec, but I scrapped nothing. Why?

The percent comes from a normal curve fitted to your data, not from a part count. It is an estimate of the long-run rate.

With a small sample you may not have made a bad part yet. The estimate is warning you that one is coming.

Can I use this tool for pass/fail or go/no-go data?

No. Cpk needs real measured values, like 50.02 mm.

For pass/fail counts, use a defect rate or a p-chart instead.

My tolerance is not symmetric, like +0.3 / −0.1. Can I still use this?

Yes. Pick Two-Sided and type the actual USL and LSL.

For a 50.0 nominal with +0.3 / −0.1, enter USL = 50.3 and LSL = 49.9. Cpk handles uneven limits fine because it checks each side on its own.

What is the difference between Cpk and Cpm?

Cpk measures distance to the nearest spec limit. Cpm also punishes you for drifting away from a stated target value, even if you are still inside spec.

This tool reports Cpk. Use Cpm when hitting nominal matters as much as staying in spec.

Why does the Target Planner show a different sigma when my mean is off-center?

Cpk is set by the closer spec limit. If your mean leans toward the USL, that side is the binding one.

The planner uses that shorter distance, so an off-center mean demands a tighter sigma to reach the same Cpk goal.

Do outliers hurt my Cpk?

Yes, a lot. One bad reading stretches the standard deviation and drags Cpk down.

Do not just delete odd points. Find out why they happened. If they came from a real process problem, they belong in the study.

Is my data saved or sent anywhere?

No. All math runs in your browser. Nothing you type is uploaded or stored.

Copy the results or take a screenshot before you close the page.

Why does the tool show Cpu or Cpl instead of Cpk for one-sided specs?

With only one limit there is no second side to compare against, so there is no Cp and no minimum to take.

The one-sided index is the correct thing to report. Many customers still call it Cpk on a report, and that is fine.

What if my data is skewed and not bell-shaped?

Cpk assumes a normal shape. Skewed data, like flatness, roundness, or runout, will give a wrong defect estimate.

Check the histogram against the fitted curve. If they do not match, use a non-normal capability method or transform the data first.