Cp and Cpk Without the Jargon: Can Your Process Actually Hit the Target?
Process capability sounds intimidating, but the question behind it is simple: given how much your process naturally wobbles, can it reliably produce output inside the limits the customer asked for? Cp and Cpk are the two numbers that answer that question. This is a plain explainer for anyone who keeps nodding along in capability discussions without being entirely sure what the indices mean.
The timing matters. After the disruptions of 2020, many operations brought work back from over-stretched suppliers or stood up new lines closer to home, often with crews still working in shifts and partly remote handovers. A new or relocated process needs to prove it can meet spec before you stake delivery promises on it — and that proof is exactly what a capability study provides.
The two ingredients: spread and centring
Capability compares two things: how wide your process varies versus how wide the specification allows. The specification has a Lower Spec Limit (LSL) and an Upper Spec Limit (USL). Your process has a mean and a standard deviation. Cp and Cpk both relate the two, but they answer slightly different questions.
Cp asks: is the process spread narrow enough to fit inside the spec width at all? It is the spec width divided by six standard deviations (USL − LSL) ÷ 6σ.
Cp ignores where the process is centred. A process can have a great Cp and still produce scrap if it drifts off to one side.
Cpk asks: given where the process is actually centred, how much room is left to the nearest spec limit? It takes the smaller of the distances to USL and to LSL, each divided by three standard deviations.
When the process is perfectly centred, Cp and Cpk are equal. The gap between them tells you how far off-centre you are.
Reading the numbers
A capability index of 1.0 means the process spread (or the nearer half of it, for Cpk) just fills the available room — there is essentially no margin. Most organizations want more breathing space than that.
Cpk below 1.0. The process is producing, or is on the edge of producing, output beyond the spec limits. You will see defects. Do not promise this process can hold tolerance.
Cpk around 1.33. A common minimum target for an established process. It corresponds to roughly four standard deviations between the mean and the nearer limit — comfortable, not heroic.
Cpk around 1.67 or higher. Often expected for safety-critical or high-volume characteristics where even small defect rates are costly.
Cp much higher than Cpk. Your variation is fine but your aim is off. Re-centre the process — that is usually a faster, cheaper fix than reducing variation.
Before you trust the index
Two cautions save a lot of embarrassment. First, capability only means something if the process is stable — in statistical control — over the study period. Run a control chart first; a Cpk computed on an out-of-control process is a number describing chaos, not capability. Second, the data must be measured well. If your gauge cannot tell good parts from bad, a measurement systems analysis comes before any capability study. Garbage in, confident-looking garbage out.
Used in order — stable process, trustworthy measurement, then Cp and Cpk — these indices turn a vague worry ("can we hold this tolerance?") into a defensible answer you can put in front of a customer or an auditor.
If you are bringing new lines online or tightening quality on existing ones and want help building capability studies that hold up to scrutiny, XNM's strategic advisory can help you put the right measures and disciplines in place.