Cp / Cpk Process Capability

Capability assumes a stable process and an appropriate distribution model.

This selection changes review context only. Formulas and example inputs stay unchanged. Check the model assumptions and applicability for the actual process.

Identify the actual asset, process, work boundary, source documents, units and assumptions. Confirm that the selected model or template applies before using its results.

1 Inputs

Minimum: 0.0001

2 Results

Cpk
1.25
Cp (potential)
1.39
Cpu (upper)
1.25
Cpl (lower)
1.53
Cp = (USL − LSL)/6σ; Cpk = min(USL−μ, μ−LSL)/3σ
Capability assumes a stable process and an appropriate distribution model. Use within-subgroup sigma for Cp/Cpk; overall sigma gives Pp/Ppk instead. Compare against the customer specification, then investigate centring and variation separately.

Worked example

Example values are illustrative. Replace them with your measured data.

InputValue
Upper spec limit (USL)10.5
Lower spec limit (LSL)9.5
Process mean (x̄)10.05
Process std dev (σ)0.12

Cpk: 1.25

References

Related tools

About this calculator

How it works

You enter the upper and lower specification limits (USL, LSL), the process mean and the process standard deviation σ. The tool computes Cp = (USL − LSL) ÷ 6σ, Cpu = (USL − mean) ÷ 3σ, Cpl = (mean − LSL) ÷ 3σ, and Cpk = the smaller of Cpu and Cpl.

Worked example (illustrative inputs)

USL 10.5, LSL 9.5, mean 10.05, σ 0.12. Cp = 1.0 ÷ 0.72 = 1.39. Cpu = 0.45 ÷ 0.36 = 1.25. Cpl = 0.55 ÷ 0.36 = 1.53. Cpk = 1.25, which the tool rates as marginal.

How to read the result

Rating bands: 1.67 or more is excellent, 1.33 or more is capable, 1.0–1.33 is marginal, and below 1.0 is not capable. When Cp is close to Cpk (within 0.05), the process is centred and only reducing σ will help. When Cp is clearly higher than Cpk, as in this example, moving the mean towards the middle of the specification is usually the quickest gain.

When to use it

Use it with σ estimated from a stable process, for example from control-chart data, to judge whether a characteristic reliably falls inside its specification.

À propos de ce calculateur

Traduction (brouillon) — à relire.

Fonctionnement

Vous saisissez les limites de spécification haute et basse (USL, LSL), la moyenne et l'écart-type σ du procédé. L'outil calcule Cp = (USL − LSL) ÷ 6σ, Cpu = (USL − moyenne) ÷ 3σ, Cpl = (moyenne − LSL) ÷ 3σ, et Cpk = le plus petit de Cpu et Cpl.

Exemple chiffré (valeurs d'illustration)

USL 10,5, LSL 9,5, moyenne 10,05, σ 0,12. Cp = 1,0 ÷ 0,72 = 1,39. Cpu = 0,45 ÷ 0,36 = 1,25. Cpl = 0,55 ÷ 0,36 = 1,53. Cpk = 1,25, que l'outil classe comme limite.

Lire le résultat

Seuils : 1,67 ou plus excellent, 1,33 ou plus capable, 1,0 à 1,33 limite, moins de 1,0 non capable. Si Cp est proche de Cpk (écart inférieur à 0,05), le procédé est centré et seule la baisse de σ aidera. Si Cp est nettement supérieur à Cpk, comme ici, recentrer la moyenne au milieu de la tolérance est en général le gain le plus rapide.

Quand l'utiliser

À utiliser avec un σ estimé sur un procédé stable, par exemple à partir de cartes de contrôle, pour juger si une caractéristique reste de façon fiable dans sa tolérance.

حول هذه الحاسبة

ترجمة (مسودة) — للمراجعة.

طريقة الحساب

تُدخل حد المواصفة الأعلى والأدنى (USL وLSL) ومتوسط العملية وانحرافها المعياري σ. تحسب الأداة Cp = (USL − LSL) ÷ 6σ، وCpu = (USL − المتوسط) ÷ 3σ، وCpl = (المتوسط − LSL) ÷ 3σ، وCpk = القيمة الأصغر بين Cpu وCpl.

مثال محسوب (قيم توضيحية)

USL = 10.5، LSL = 9.5، المتوسط 10.05، σ = 0.12. Cp = 1.0 ÷ 0.72 = 1.39. Cpu = 0.45 ÷ 0.36 = 1.25. Cpl = 0.55 ÷ 0.36 = 1.53. Cpk = 1.25، وتصنفه الأداة كهامشي.

قراءة النتيجة

التصنيف: 1.67 فأكثر ممتاز، 1.33 فأكثر قادر، من 1.0 إلى 1.33 هامشي، وأقل من 1.0 غير قادر. إذا كان Cp قريباً من Cpk (فرق أقل من 0.05) فالعملية متمركزة ولا يفيد إلا خفض σ. وإذا كان Cp أعلى بوضوح من Cpk، كما في هذا المثال، فإن نقل المتوسط إلى منتصف المواصفة هو عادة أسرع تحسين.

متى تستخدمه

استخدمها مع σ مقدّر من عملية مستقرة، مثلاً من بيانات خرائط التحكم، للحكم على ما إذا كانت الخاصية تقع بثبات داخل مواصفتها.

In one line

Capability assumes a stable process and an appropriate distribution model. Use within-subgroup sigma for Cp/Cpk; overall sigma gives Pp/Ppk instead. Compare against the customer specification, then investigate centring and variation separately.

Worked example

EXAMPLE

Illustrative values (the calculator defaults), not data from a real site.

Inputs
Upper spec limit (USL)10.5
Lower spec limit (LSL)9.5
Process mean (x̄)10.05
Process std dev (σ)0.12
Results
Cpk1.25
Cp (potential)1.39
Cpu (upper)1.25
Cpl (lower)1.53
CapabilityNot yet

Frequently asked questions

What's the difference between Cp and Cpk, and why do they sometimes differ a lot?

Cp assumes the process is perfectly centred between the spec limits, while Cpk uses the actual mean, so it penalises a process that's drifted toward one limit. A big gap between Cp and Cpk means the process has plenty of potential capability but is off-centre — recentring the mean is usually the quickest fix in that case.

What Cpk value should I be aiming for?

1.33 is the widely-used minimum for an established process; 1.67 or higher is often required for safety-critical or new processes; 2.0 corresponds to true Six Sigma performance. Below 1.0, the process is statistically expected to produce output outside spec even when it's behaving normally.

Where do I get the process standard deviation?

Ideally from a control-chart study — for example estimating sigma as R-bar/d2 from an X-bar and R chart's average range — using data collected while the process is in statistical control. Using sigma from an out-of-control or unrepresentative dataset will give a misleading capability number.

Can a process be in control but still have a poor Cpk?

Yes — being in statistical control means the process is stable and predictable, which is a separate question from whether that stable output actually fits inside the customer's spec limits. A perfectly stable process can still have a low Cpk if its spread is too wide or its mean sits too close to a limit.

Why can't I just improve Cpk by widening the spec limits?

You can mathematically, but the spec limits represent what the customer or design actually requires — widening them to make the number look better doesn't change whether the parts function correctly. Real capability improvement comes from centring the process and/or reducing variation, not moving the goalposts.

Donate · Faire un don · تبرّع