Cpk is the process capability index—a statistical measure of how well a process output fits within its specification limits while accounting for how centered the process is. It compares the process mean and variation with the nearest specification limit. In many manufacturing applications, Cpk ≥ 1.33 is a commonly used capability benchmark.
Cpk, or process capability index, measures how consistently a manufacturing process produces output within specified limits, with higher values indicating better quality and less variation. Calculating Cpk involves comparing the process mean and standard deviation to the upper and lower specification limits, where a value above 1.0 signifies a capable process.
Cpk is widely used in manufacturing, quality engineering, Six Sigma, and process improvement because it answers a practical question: Can this process consistently produce output within the customer’s or engineering specification limits?
Unlike a simple measure of variation, Cpk considers both the width of the process distribution and its location relative to the upper and lower specification limits. A process may have relatively little variation but still perform poorly if its mean has shifted too close to one specification limit.
That makes Cpk particularly valuable for identifying processes that appear capable on paper but are not adequately centered in practice.
What Is Cpk?
Cpk, or the process capability index, measures the capability of a process relative to its specification limits while accounting for both process variation and process centering.
In plain English, Cpk tells you how much room your process has between its average output and the closest specification limit, relative to the amount of natural process variation.
A higher Cpk generally indicates that the process distribution fits more comfortably within the specification limits. A lower Cpk indicates that the process has less margin for variation and may produce a greater proportion of output beyond specifications.
Cpk considers three major elements:
- USL: Upper Specification Limit
- LSL: Lower Specification Limit
- Process mean (μ): Where the process is centered
- Standard deviation (σ): How much the process varies
Specification limits are determined by customer, engineering, design, regulatory, or other requirements. They are not the same as control limits, which are calculated from process data and describe expected process behavior.
Why Does Cpk Matter?
Imagine a shaft with an acceptable diameter between 9.90 mm and 10.10 mm.
A process centered at exactly 10.00 mm with relatively little variation has substantial room on both sides before reaching either specification limit.
Now imagine the same amount of variation, but the process mean shifts to 10.07 mm.
The spread has not changed, yet the process is much closer to the upper specification limit. Even ordinary process variation can now result in more out-of-specification parts.
Cpk detects that problem because it evaluates the distance from the mean to the nearest specification limit.
Why Cpk Can Tell You More Than Cp
Cp measures potential capability based primarily on specification width relative to process spread.
Cpk goes further by considering where the process mean actually sits.
If a process is perfectly centered, Cp and Cpk may be equal or nearly equal.
If the process is off-center:
Cpk < Cp
The larger the difference between Cp and Cpk, the stronger the indication that centering is reducing actual capability.
This is why Cpk is often especially useful when evaluating real process performance rather than theoretical capability alone.
How Cpk Connects to Six Sigma Quality Levels
Cpk and Six Sigma are closely related because both evaluate process variation relative to specification limits. However, the connection needs to be interpreted carefully.
Cpk measures the distance between the process mean and its nearest specification limit in units of three standard deviations:
Distance to nearest specification limit ≈ 3 × Cpk × σ
This creates a useful relationship between Cpk and short-term sigma distance:
| Cpk | Approximate Sigma Distance to Nearest Spec | General Interpretation |
|---|---|---|
| 1.00 | 3σ | Limited capability margin |
| 1.33 | ≈ 4σ | Common manufacturing capability benchmark |
| 1.67 | ≈ 5σ | Higher capability margin |
| 2.00 | 6σ | Six-sigma distance under the centered short-term model |
For example:
3 × 1.33 ≈ 4σ
This means the nearest specification limit is approximately four standard deviations from the process mean.
Similarly:
Cpk = 1.67 → approximately 5σ
and:
Cpk = 2.00 → 6σ
This is why a perfectly centered process with Cpk = 2.0 is often described as having Six Sigma short-term process capability.
However, this should not automatically be interpreted as the famous 3.4 defects per million opportunities (DPMO) associated with Six Sigma.
The traditional Six Sigma 3.4 DPMO convention incorporates an assumed 1.5σ long-term shift in the process mean. A centered normal process with specification limits six standard deviations from the mean has a much lower theoretical defect rate if no such shift occurs.
Therefore, it is important to distinguish between:
Cpk-to-sigma distance: a mathematical relationship describing how many standard deviations separate the process mean from its nearest specification limit.
Six Sigma DPMO conventions: a quality-performance convention that may incorporate assumptions about long-term process movement.
Cpk is especially useful within Six Sigma projects because it provides a quantitative way to assess process capability before and after improvement. During a DMAIC project, teams may calculate baseline Cpk during the Measure phase, identify sources of variation or poor centering during Analyze, improve the process during Improve, and continue monitoring capability during Control.
Increasing Cpk generally requires one or both of two improvements: reduce process variation or move the process mean closer to its optimal target. Both goals align directly with Six Sigma’s broader objective of creating more predictable processes and reducing defects.
Cpk should therefore be viewed as an important capability metric within Six Sigma—not as a complete substitute for defect analysis, process stability, control charts, or the broader DMAIC methodology.
The Cpk Formula
The standard Cpk formula is:
Cpk = min[(USL − μ) / 3σ, (μ − LSL) / 3σ]
The two terms can also be calculated separately:
Cpu = (USL − μ) / 3σ
Cpl = (μ − LSL) / 3σ
Then:
Cpk = min(Cpu, Cpl)
where:
- Cpk = process capability index
- USL = upper specification limit
- LSL = lower specification limit
- μ = process mean
- σ = estimated within-process standard deviation
- Cpu = upper capability index
- Cpl = lower capability index
Cpk uses the smaller of Cpu and Cpl because the closest specification limit presents the greatest immediate risk.
Why Is Standard Deviation Multiplied by 3?
For an approximately normal distribution, three standard deviations from the mean cover one side of the familiar ±3σ range.
The total natural process spread is often described as approximately:
6σ
or:
μ − 3σ to μ + 3σ
Cpk evaluates how many three-standard-deviation distances fit between the process mean and its nearest specification limit.
Worked Cpk Example
Suppose a manufacturing process produces a component with these specifications:
LSL = 9.80 mm
USL = 10.20 mm
Process data indicates:
μ = 10.05 mm
σ = 0.04 mm
First calculate the upper capability:
Cpu = (10.20 − 10.05) / (3 × 0.04)
Cpu = 0.15 / 0.12
Cpu = 1.25
Now calculate the lower capability:
Cpl = (10.05 − 9.80) / (3 × 0.04)
Cpl = 0.25 / 0.12
Cpl ≈ 2.08
Cpk is the smaller result:
Cpk = min(1.25, 2.08)
Cpk = 1.25
Although the process has considerable room above the lower specification, it is much closer to the upper specification. The upper side therefore limits its capability.
This example demonstrates one of Cpk’s greatest strengths: it identifies which specification limit is constraining the process.
Cp vs. Cpk
Cp and Cpk are closely related capability indices, but they answer different questions.
Cp asks: How well could the process fit within the specifications if it were properly centered?
Cpk asks: How well does the process fit when its actual centering is considered?
The Cp formula is:
Cp = (USL − LSL) / 6σ
Cp compares total specification width with total process spread.
Cpk instead evaluates the mean’s distance from each specification limit and selects the smaller result.
| Measure | Cp | Cpk |
| Measures variation vs. specifications | Yes | Yes |
| Accounts for process centering | No | Yes |
| Uses both specification limits directly | Yes | Yes |
| Identifies nearest-limit constraint | No | Yes |
| Represents | Potential capability | Centering-adjusted capability |
If:
Cp ≈ Cpk
the process is likely reasonably centered.
If:
Cp > Cpk
the process may have adequate potential capability but is losing capability because its mean is shifted.
For a complete explanation, see Cp vs. Cpk.
How to Calculate Cpk
A Cpk calculation requires more than inserting numbers into a formula. The process data should first be suitable for capability analysis.
At a high level:
- Define the characteristic being measured.
- Identify the correct USL and LSL.
- Collect representative process data.
- Verify that the measurement system is adequate.
- Confirm that the process is reasonably stable.
- Evaluate the distribution and assumptions.
- Calculate the process mean.
- Estimate within-process standard deviation.
- Calculate Cpu and Cpl.
- Use the smaller result as Cpk.
For example, if:
USL = 50
LSL = 40
μ = 46
σ = 1
then:
Cpu = (50 − 46) / 3 = 1.33
and:
Cpl = (46 − 40) / 3 = 2.00
Therefore:
Cpk = 1.33
The upper specification is the limiting side.
For a complete worked walkthrough, see How to Calculate Cpk.
Cpk Value Ranges: What Is a Good Cpk?
A common question in capability analysis is: What is a good Cpk value?
There is no universal threshold appropriate for every industry, customer, characteristic, or risk level. Organizations should follow applicable customer, regulatory, engineering, or internal requirements.
However, these values are frequently used as practical reference points:
| Cpk Value | General Interpretation |
| < 1.00 | Process spread/centering does not fit adequately within specifications |
| 1.00–1.33 | Often considered marginal or requiring improvement |
| 1.33 | Commonly used minimum capability benchmark |
| 1.33–1.67 | Generally capable under many conventional requirements |
| 1.67 | Frequently associated with higher capability requirements |
| 2.00 | Very high capability under the conventional short-term normal model |
What Does Cpk = 1.0 Mean?
A Cpk of 1.0 means the nearest specification limit is approximately three within-process standard deviations from the mean.
Under an ideal normal and stable-process model, the process just fits to the closest specification boundary at approximately the 3σ distance.
That leaves relatively little margin for process shifts.
What Does Cpk = 1.33 Mean?
A Cpk of 1.33 places the closest specification limit approximately four standard deviations from the process mean:
1.33 × 3 ≈ 4σ
This additional margin is one reason 1.33 is frequently used as a conventional minimum for an established manufacturing process.
It should not, however, be treated as a universal law.
What Does Cpk = 1.67 Mean?
A Cpk of approximately 1.67 corresponds to about five standard deviations between the mean and the nearest specification limit.
Organizations may require this higher level for new processes, important characteristics, or applications where additional capability margin is desirable.
What Does Cpk = 2.0 Mean?
Cpk = 2.0 corresponds to six within-process standard deviations between the mean and the nearest specification limit when the process is centered appropriately.
It is sometimes described as a Six Sigma capability level under the short-term, centered normal model. This should not be confused mechanically with long-term Six Sigma defect-rate conventions, which may incorporate an assumed process shift.
For a deeper discussion of capability thresholds, see Cpk Values: What Is a Good Cpk?.
Cpk in Manufacturing
Cpk is particularly important in manufacturing because specifications determine whether output satisfies design and customer requirements.
A manufacturer might calculate Cpk for:
- Shaft diameter
- Part thickness
- Hole diameter
- Torque
- Fill weight
- Tensile strength
- Flatness
- Temperature
- Chemical concentration
- Assembly dimensions
A high capability index provides evidence that ordinary process variation fits comfortably within the applicable specification limits.
A low Cpk indicates that the process mean is too close to a limit, process variation is too large, or both.
Why 1.33 Is a Common Manufacturing Benchmark
A Cpk of 1.00 leaves the closest specification boundary at about three standard deviations from the process mean.
Manufacturers often want additional margin because real processes can drift.
At Cpk = 1.33, the nearest limit is approximately four standard deviations away. That extra margin makes the process more tolerant of modest variation or movement while remaining within specification.
For that reason, 1.33 is widely used as a conventional minimum benchmark for many established manufacturing processes.
However, it is not automatically the correct requirement for every characteristic.
Customer-specific requirements, control plans, PPAP requirements, internal standards, safety considerations, and the criticality of the characteristic can require different thresholds.
Automotive and AIAG Context
Process capability studies are widely used in automotive quality systems and related supplier requirements. Capability expectations may be incorporated into product and process validation, production approval, control planning, and ongoing monitoring.
Rather than assuming that every automotive application uses the same Cpk threshold, quality teams should verify the current customer-specific requirement and applicable AIAG or OEM methodology.
Some organizations apply higher capability expectations to new processes, special characteristics, or characteristics associated with greater product risk.
The practical lesson is:
Treat 1.33 as a common reference point—not as a substitute for the actual requirement governing your process.
Capability Analysis: How to Run a Cpk Study
A meaningful capability analysis begins before Cpk is calculated.
Software can produce a Cpk number almost instantly. The difficult part is ensuring that the number describes a process appropriately.
Real-World Manufacturing Cpk Example (With Data)
Consider a manufacturer producing precision-machined shafts with a required diameter between 24.90 mm and 25.10 mm. To evaluate whether the machining process can consistently meet these specifications, the quality team collects the following 20 measurements:
25.01, 25.03, 25.00, 25.04, 25.02, 25.05, 25.01, 25.02, 25.04, 25.03, 25.00, 25.02, 25.05, 25.01, 25.03, 25.04, 25.02, 25.01, 25.03, 25.02
For this example:
| Measure | Value |
|---|---|
| Lower Specification Limit (LSL) | 24.90 mm |
| Upper Specification Limit (USL) | 25.10 mm |
| Process mean | 25.024 mm |
| Sample standard deviation | ≈ 0.0150 mm |
| Cpu | ≈ 1.69 |
| Cpl | ≈ 2.75 |
| Cpk | ≈ 1.69 |
Using the Cpk formulas:
Cpu = (USL − μ) / 3σ
Cpu = (25.10 − 25.024) / (3 × 0.0150) ≈ 1.69
For the lower specification limit:
Cpl = (μ − LSL) / 3σ
Cpl = (25.024 − 24.90) / (3 × 0.0150) ≈ 2.75
Because Cpk uses the smaller value:
Cpk = min(1.69, 2.75) = 1.69
A Cpk of approximately 1.69 indicates substantial capability margin relative to the specifications in this example. However, the calculation also reveals something important: the upper specification limit is much closer to the process mean than the lower limit.
The process therefore has more capability on the lower side than the upper side. If the process mean could be moved closer to the nominal target of 25.00 mm without increasing variation, Cpk could improve further.
This demonstrates why manufacturers should look beyond the final Cpk number. Comparing Cpu and Cpl reveals which side of the specification is limiting capability and can help determine whether improvement should focus on process centering, variation reduction, or both.
Before treating any Cpk result as evidence of future capability, manufacturers should also verify that the measurement system is adequate, the data are representative, the process is reasonably stable, and the selected capability model is appropriate for the data.
Define the Process and Characteristic
Clearly identify what you are measuring.
A capability study should address a specific measurable characteristic with meaningful specification limits.
For example:
Characteristic: machined shaft diameter
LSL: 24.90 mm
USL: 25.10 mm
Do not substitute control limits for specification limits.
Verify the Measurement System
If the measurement system introduces excessive error, the calculated process variation may not represent the actual process.
Depending on the application, a measurement system analysis such as Gage R&R may be appropriate before capability is assessed.
Poor measurement can make a capable process appear incapable—or hide genuine process variation.
Collect Representative Data
The data should represent the process conditions you want the Cpk study to describe.
Consider factors such as:
- Shifts
- Machines
- Operators
- Materials
- Tools
- Lots
- Environmental conditions
- Production periods
A large quantity of biased or unrepresentative data does not create a reliable capability study.
Check Process Stability
Capability and stability are different concepts.
A process can be within specification today but still be statistically unstable.
Control charts can help determine whether special-cause variation is present. If the process is changing unpredictably, a single capability index may provide a misleading description of future performance.
In general, establish reasonable statistical control before relying heavily on Cpk.
Check the Distribution
Conventional Cpk analysis commonly relies on assumptions associated with an approximately normal distribution.
Inspect the data with tools such as:
- Histogram
- Probability plot
- Normality assessment
- Box plot
- Time-series plot
If the data is strongly skewed or follows a known non-normal distribution, forcing a conventional normal Cpk calculation may produce misleading results.
Statistical software can perform non-normal capability analysis using suitable distributions or transformation methods.
Calculate and Interpret Cpk
Once the process and data are suitable, calculate Cpu and Cpl.
Do not stop at the final Cpk number.
Ask:
Which side limits capability?
Is the problem excessive variation, poor centering, or both?
How does Cpk compare with Cp?
Does the process meet the actual customer or engineering requirement?
These questions turn capability analysis into actionable process information.
When Should You Use Cpk vs. Ppk?
Cpk and Ppk both compare process output with specification limits, but they commonly use different estimates of variation.
Cpk typically uses an estimate of within-subgroup or short-term variation.
Ppk uses the overall standard deviation observed across the full dataset.
As a result, Ppk can reflect additional sources of variation that occur over time.
If the process is stable and consistent, Cpk and Ppk may be relatively close.
If:
Cpk is substantially higher than Ppk
the process may be experiencing shifts, drift, batch-to-batch differences, or other longer-term variation that is not captured fully by the within-subgroup estimate.
This difference can itself provide valuable diagnostic information.
Cpk should therefore not automatically replace Ppk, nor should the two indices be treated as interchangeable.
Common Mistakes With Cpk
One of the most serious mistakes is calculating Cpk for an unstable process and treating the result as evidence of predictable future capability.
Capability assumes that the process being described has meaningful consistency.
Another common error is confusing specification limits with control limits.
Specification limits come from requirements.
Control limits come from process behavior.
They serve different purposes.
Other frequent mistakes include:
- Assuming the data is normal without checking
- Using nonrepresentative samples
- Ignoring measurement-system error
- Reporting Cpk without Cp
- Ignoring the difference between Cpu and Cpl
- Treating 1.33 as a universal requirement
- Assuming a high Cpk proves the process is stable
- Confusing Cpk with Ppk
- Calculating capability from too little data
- Rounding intermediate calculations too aggressively
Perhaps the biggest conceptual mistake is treating Cpk as a score that exists independently of the process.
A Cpk value is meaningful only in relation to the process, data collection method, variation estimate, specification limits, distribution, and time period used to calculate it.
Using Cpk for Process Improvement
Cpk is most useful when it leads to action.
Suppose Cp is high but Cpk is considerably lower. The process has potentially adequate spread but is not centered well.
The improvement priority may be to shift the process mean toward the target.
If both Cp and Cpk are low, centering alone may not solve the problem. The process may require a reduction in variation through changes to equipment, tooling, materials, methods, measurement, environmental controls, or other process inputs.
This creates a useful diagnostic pattern:
High Cp + lower Cpk → investigate centering
Low Cp + low Cpk → investigate variation
High Cp + high Cpk → process has strong potential and centering-adjusted capability
Even a capable process still requires control. Capability analysis does not replace ongoing monitoring.
Common Manufacturing Scenarios Where Cpk Fails
Cpk is a powerful process capability metric, but it can produce misleading conclusions when it is applied to the wrong data or process conditions. A Cpk calculation is only as meaningful as the process and assumptions behind it.
The process is statistically unstable
One of the most important problems occurs when a process contains special-cause variation. Tool wear, equipment failures, material changes, incorrect machine settings, operator differences, or environmental changes may cause the process mean or variation to shift over time.
A single Cpk value can hide these changes.
For example, a process might have an acceptable overall Cpk while its mean is steadily drifting toward the upper specification limit. A control chart may reveal the instability that the capability number alone does not show.
Whenever possible, establish reasonable statistical control before relying heavily on Cpk as a predictor of future performance.
The data are strongly non-normal
Traditional Cpk calculations are commonly interpreted using an approximately normal process model. If the actual distribution is strongly skewed, bounded, multimodal, or otherwise non-normal, a conventional normal Cpk calculation can misrepresent the process.
In these situations, quality professionals may need an appropriate non-normal capability analysis, distribution model, or transformation rather than forcing the data into a normal assumption.
Different production streams are pooled together
Suppose measurements from several machines, cavities, production lines, suppliers, or shifts are combined into one dataset. The resulting distribution may contain multiple underlying process populations.
A single Cpk value can then hide an incapable machine or shift behind the stronger performance of another.
Stratifying the data by machine, tool, cavity, material lot, operator, or other meaningful process factors can reveal variation that an aggregated capability calculation conceals.
The measurement system is inadequate
Cpk depends on the observed process variation. If the measurement system contributes excessive error, the calculated standard deviation may not accurately represent the manufacturing process.
A Gage R&R or other measurement system analysis may therefore be necessary before performing a formal capability study. Poor measurement can make a good process appear worse—or obscure genuine process problems.
Specification limits are confused with control limits
Specification limits and control limits answer different questions.
Specification limits define acceptable product requirements and normally come from customers, engineering, design, or regulatory requirements.
Control limits are calculated from process data and describe expected process behavior.
Using control limits as USL and LSL values produces a capability calculation that does not answer whether the process meets the actual product specification.
The sample is too small or unrepresentative
A Cpk calculated from a short production run under ideal conditions may look excellent but fail to represent normal production.
Capability data should reflect the operating conditions the study is intended to characterize, including relevant machines, shifts, operators, materials, tooling, and environmental conditions.
Cpk and Ppk are treated as interchangeable
Cpk commonly uses a within-process estimate of variation, whereas Ppk uses the overall variation observed across the dataset.
If Cpk is much higher than Ppk, the process may be experiencing longer-term shifts, drift, batch differences, or other sources of variation. Reporting only the more favorable number can create an incomplete picture of actual process performance.
The key lesson is that Cpk does not fail because the formula is wrong—it fails when the calculation is used without understanding the process behind the number.
Turning Manufacturing Capability Results Into Action
Connect the capability result to the production characteristic you can control. For a machined diameter, separate measurements by machine, cavity or setup before deciding whether the process needs recentering. For a filling process, examine dosing settings, material conditions and changes between shifts. For an assembly characteristic, check both the measurement system and the upstream inputs.
If the process is stable and its mean is too near a specification limit, investigate centering. If its spread is too wide, investigate the sources of variation. After a controlled change, collect representative data and repeat the study rather than combining unlike before-and-after observations.
For the interpretation of one common benchmark, see what a Cpk of 1.33 means. For the distinction between potential and centered capability, use the Cp vs. Cpk comparison.
Conclusion
Cpk provides a concise way to evaluate the capability of a process by comparing its mean and variation with the nearest specification limit. A Cpk of 1.33 is a common manufacturing reference point, while higher values such as 1.67 may indicate greater capability margin and may be required for certain applications. The number becomes meaningful only when it comes from representative data, an adequate measurement system, an appropriately stable process, and a suitable capability model. Comparing Cpk with Cp, Ppk, Cpu, and Cpl can reveal whether poor capability is being driven primarily by variation, centering, or longer-term process behavior. Used as part of a disciplined capability analysis rather than as an isolated score, Cpk can help quality teams identify improvement priorities, reduce nonconforming output, and build more predictable processes.
Frequently Asked Questions
What does Cpk mean in manufacturing?
In manufacturing, Cpk (Process Capability Index) measures how well a production process operates within its specification limits, accounting for both the spread of variation and where the process is centered. A Cpk of 1.33 is the widely accepted manufacturing target, meaning the process fits within specs with a margin of safety. Higher Cpk values indicate more capable, lower-defect processes. Cpk below 1.0 means the process is producing out-of-spec output.
What is the Cpk formula in manufacturing?
The Cpk formula is Cpk equals the minimum of (USL minus x-bar) divided by 3 sigma, or (x-bar minus LSL) divided by 3 sigma. USL is the upper specification limit; LSL is the lower specification limit; x-bar is the process mean; sigma is the process standard deviation. Cpk takes whichever of the two values is smaller because that side is the closer to being out-of-spec — the binding constraint.
What is a good Cpk value in manufacturing?
A Cpk of 1.33 is the industry-standard target for manufacturing processes, corresponding to approximately 4 sigma performance. A Cpk of 1.67 corresponds to about 5 sigma and is the Six Sigma capability target. A Cpk of 2.0 is world-class, corresponding to Six Sigma with a 1.5-sigma shift allowance. Automotive suppliers typically require 1.33 as a minimum and 1.67 for critical characteristics.
What is the difference between Cp and Cpk in manufacturing?
Cp measures only whether the process spread fits within specification limits; it assumes the process is centered. Cpk measures both spread and centering, penalizing a process that fits the spec width but is off-center. When Cp equals Cpk, the process is perfectly centered. When Cpk is less than Cp, the process is off-center and could be improved by centering alone. Manufacturers use Cpk for real capability assessment; Cp is used to see the ceiling if centering were fixed.
How do you calculate Cpk on a manufacturing line?
First, collect at least 30 measurements from the process under stable conditions. Second, calculate the mean (x-bar) and standard deviation (sigma) of those measurements. Third, identify the USL and LSL from the specification. Fourth, calculate both (USL – x-bar) / (3 sigma) and (x-bar – LSL) / (3 sigma). Fifth, take the minimum of the two values as Cpk. Sixth, verify the process is in statistical control before drawing conclusions.
Why does my Cpk keep failing on the manufacturing line?
Failing Cpk usually comes from four causes. First, the process is off-center (fix by centering the mean between USL and LSL). Second, the process variation is too large (fix by reducing variation through improved tooling, materials, or controls). Third, the specification is too tight for the current process capability (renegotiate spec or invest in process improvement). Fourth, the measurement system itself is contributing variation (run a Gage R&R study before blaming the process).
