What is it, and why use it?
Process Capability compares measured process variation with explicit specification limits, using indices and distribution diagnostics under stated assumptions.
It helps a team assess how a supported process model relates to an agreed requirement, beyond looking only at an average.
Explore the Process Capability Analysis features and local-file workflowWhen to use it
Use it after reviewing measurement quality, process stability, sampling and the suitability of a distribution model.
When another tool fits better
Use a Control Chart first if stability is uncertain. Use CTQC to agree the requirement instead of inventing specification limits from the observed data.
Before you start
- Representative measured values with consistent units and measurement quality.
- Explicit lower and/or upper specification limits from an agreed requirement.
- A documented distribution assessment and a separate stability review.
Use fictional or approved non-sensitive information. Do not include patient details or proprietary records in practice worksheets.
Step-by-step in CI Tools
Enter measured values and the explicit specification limits; do not manufacture a second limit for a one-sided requirement.
Inspect the histogram, distribution diagnostics, measurement-system information and stability screen.
Record a supporting distribution assessment before interpreting normal-model indices; the application withholds them until this is recorded.
Review Cp/Cpk using within variation and Pp/Ppk using overall sample standard deviation. Check one-sided availability and warnings.
Save the data, assumptions and editable analysis. Use results as conditional estimates requiring local technical review, not certification.
Check stability and distribution before interpreting indices
A numerical index can be misleading when the process shifts, measurements are unreliable or the model is inappropriate. Do not infer a missing specification limit. Report assumptions, sample context and warnings together with any estimate.
Completed worked examples
Demonstrations only: all organizations, observations, targets and results below are fictional. They are not clinical instructions or universal performance standards.
Healthcare operations — fictional
Administrative handling requirement
Starting information: A fictional service considers an operational handling-time requirement.
| Inputs | Representative times in minutes and a locally agreed upper specification |
|---|---|
| Review | Time sequence suggests a process shift; a normal-model assumption is not yet supported |
| Completed conclusion | Do not report a normal capability index; investigate stability and distribution first |
What the result means: Withholding an index is a meaningful result when assumptions are unsupported. The tool must not turn an incomplete assessment into evidence of suitability.
General business — fictional
Synthetic component length
Starting information: A fictional stable process is assessed under a supported normal model.
| Specifications / estimates | LSL 9.5 mm, USL 10.5 mm; mean 10.1 mm, within SD 0.1 mm |
|---|---|
| Illustrative Cp | (10.5 − 9.5) / (6 × 0.1) = 1.67 |
| Illustrative Cpk | min((10.5 − 10.1)/0.3, (10.1 − 9.5)/0.3) = 1.33 |
What the result means: Cpk is lower because the mean is off-centre. These illustrative values assume the stated estimates and model; they are not a universal acceptance threshold or a calculation from a real dataset.
Practise in a separate worksheet
Open the tool’s built-in synthetic sample as a new editable worksheet. It may differ from the worked demonstrations above. It never replaces an existing worksheet. Save any work you want to retain to your computer.
Open built-in practice worksheetYou’ll be asked to log in if needed. No worksheet is created just by reading this page.
Common mistakes and how to avoid them
- Using limits estimated from the same sample as customer specifications: agree requirements independently.
- Assuming any histogram is normal: record and review distribution support.
- Reporting Cp without centring, stability or uncertainty: review the full evidence.
Interpret the result
Indices are conditional on the data, spread estimate and model. Cp describes potential spread relative to two-sided limits; Cpk also reflects centring. Overall indices use a different spread estimate. None establishes clinical acceptability or replaces specialist review.
Record the assumptions and unanswered questions alongside the result. Use the relevant local review process before making operational decisions.
Credible further learning
These independent educational resources explain the methods. They do not endorse CI Tools Suite. Some external resources have their own access requirements.
- NIST: Process capability (external resource)
- NIST: Process monitoring and control (external resource)
Resource links reviewed 29 October 2026. Our interface instructions and fictional examples are original CI Tools learning material.
Put the method into practice
Start a new worksheet, use the guide beside your work and keep an editable .citools copy on your computer.
Try This ToolChoose another tool