Cpk is a measurement of process performance, not process potential. The distinction sounds technical but has practical implications for how precision manufacturers use capability data and what decisions they can legitimately make from it.
A Cpk of 1.67 tells you that with your current parameter settings, your current tooling, and your current material lot, the characteristic you measured has a distribution centered close enough to the specification mean and narrow enough to put approximately 0.6 defects per million opportunities outside the specification limits. This is useful. It tells you whether your current configuration can hold the spec.
What it does not tell you is whether a different parameter configuration would deliver a Cpk of 2.0 or higher, shifting the defect rate lower and increasing your margin against process drift. That question is not answerable from capability data alone. It requires exploring the parameter space.
What Cpk Measures and Where the Confusion Comes From
Cpk equals the minimum of (USL minus x-bar) divided by 3-sigma, and (x-bar minus LSL) divided by 3-sigma, where USL and LSL are the upper and lower specification limits, x-bar is the process mean, and sigma is the within-subgroup standard deviation. Higher is better: Cpk 1.33 is typically the minimum acceptable for a controlled production process, 1.67 is preferred for critical characteristics, and 2.0 or above is required for some safety-critical aerospace and defense applications.
The confusion arises because Cpk summarizes all sources of within-subgroup variation into a single number without distinguishing between them. Two processes can have identical Cpk values but very different sources of variation: one may have high short-term repeatability but poor centering, another may be well-centered but have high tool-wear-driven drift over a production run. The Cpk number is the same, but the improvement paths are completely different.
More fundamentally, Cpk says nothing about the shape of the parameter response surface around the current operating point. A process running at Cpk 1.33 might be sitting in a flat region of the parameter space where small adjustments change Cpk only marginally in any direction. Or it might be sitting near the edge of a high-performance region where adjusting two parameters by modest amounts would push Cpk to 1.8 or above. Capability data alone cannot distinguish between these two situations.
The Incremental Improvement Trap
Shops that run capability studies and find marginal Cpk numbers typically respond by tightening process controls: tighter tool change intervals, more frequent fixture verification, reduced material batch sizes to control lot variation. These responses reduce variability without changing the underlying parameter configuration. They can move Cpk from 1.2 to 1.4, but they cannot reliably move it from 1.4 to 1.8 if the current parameter set is fundamentally sitting in a suboptimal region.
The incremental improvement trap occurs when a shop repeatedly invests in control tightening and gets diminishing returns, without recognizing that the ceiling is the parameter configuration, not the process controls. Once controls are already tight, further tightening yields less Cpk improvement per dollar of effort invested. The leverage point shifts to the parameter level.
We are not suggesting that statistical process control is unnecessary. For a well-characterized process at good Cpk, SPC is the right ongoing investment. The problem is when SPC becomes the primary response to a marginal Cpk, because it addresses within-parameter-setting variability rather than the possibility that better settings exist.
What a Capability Study Does Well
Cpk studies are well-suited to characterizing a process that is already near its potential: confirming that a recent parameter change improved capability, certifying a production line before first article approval, monitoring for process drift over time, and building the historical record that customers and auditors require.
For AS9100-registered shops and defense sub-tier suppliers, capability documentation is mandatory for many characteristics. A Cpk study run at the right time in the right way generates both compliance documentation and genuine process intelligence. The compliance value and the process intelligence value are not in conflict, but they serve different purposes.
The intelligence value is in what the Cpk trend tells you about process stability over time, not in any single Cpk number as a measure of potential. If Cpk on a critical bore dimension has been slowly declining from 1.8 to 1.55 over six months of production, that trend is actionable: it suggests progressive tool wear effects, material lot drift, or fixture degradation, all of which can be investigated and addressed. A declining trend at a high starting Cpk is very different from a flat trend at a marginal Cpk.
Parameter Search as the Complement to Capability Studies
The questions that capability studies answer and the questions that parameter search answers are different, and both sets of questions matter.
Capability studies answer: are my current settings performing to specification? Are they stable? Are they meeting the customer's Cpk requirements? Is this process under control?
Parameter search answers: do better settings exist in the feasible parameter space? What would a higher-capability configuration look like? How much unexplored potential remains?
These questions belong in sequence. Parameter search is most valuable when a process is near or below a target Cpk, as a tool for identifying configurations that might deliver better performance. Once a better configuration is found and stabilized, capability studies confirm and document the improvement. The cycle repeats when Cpk begins drifting again or when a design revision changes the specifications.
A precision shop running titanium brackets for an aerospace defense program provides a useful example of this cycle. The shop had been running at Cpk 1.35 on a primary bore dimension against a customer requirement of Cpk 1.67. Controls were already tight: tool changes on a fixed interval, CMM inspection every 10 parts, fixture verification every shift. Further tightening controls was not moving the Cpk number meaningfully.
A parameter search across the joint feed rate, cutting speed, and depth-of-cut space for the finishing pass identified a configuration that the shop had not previously tried, at the intersection of a lower feed rate and a slightly higher cutting speed than their standard practice. The predicted yield improvement was modest in the model, but the actual Cpk after running the new configuration was 1.72, above the customer requirement. The shop's SPC monitoring confirmed the improvement was stable over the following production run.
Building a Process Development Practice That Uses Both
Precision shops that integrate parameter search and capability studies as a coherent practice rather than independent tools tend to spend less time in the marginal Cpk zone. The capability study tells them when a characteristic is underperforming. The parameter search tells them what to try next. The capability study after the parameter change confirms whether the new configuration delivered the expected improvement.
The integration requires that parameter records and capability records share a common identifier structure: for each capability measurement dataset, you need to know which parameter configuration was running, so that when you find a better configuration and re-run the study, you can compare directly. Without that linkage, capability studies and parameter records are two separate information pools that cannot inform each other.
Cpk is not the problem. The limitation is treating capability measurement as a destination rather than as one stage in a continuous improvement loop that includes parameter exploration as a legitimate tool.