VYRCOWhat you will learn
- Define the question and feature before collecting numbers.
- Keep time, cavity and conditions: an average cannot recover lost identities.
- The calculator produces indices; the study must justify data, assumptions and scope.
Who it is for: Quality, engineering and process owners.
State the question the study must answer
Describing an already produced batch, investigating differences between cavities and evaluating a process over a defined period are different objectives. Write the question and the decision it should support. Add part, revision, feature and unit; avoid a vague goal such as “get a Cpk”.
Record who requested the study and which requirements apply. Define exclusions: machines, cavities, shifts or conditions outside its scope. If the scope changes during the work, keep the plan revision so the report does not appear to answer a question that was never studied.
Identify specification and method sources
Specification limits belong to the product requirement. Retain drawing, revision and the authority behind any clarification. Do not use the sample minimum and maximum as specification limits. Control limits serve a different purpose and do not replace drawing requirements either.
Before collecting data, document the method, instrument and applicable measurement-system review. Displayed resolution does not describe all the information needed to evaluate a result. If a condition required by the procedure is missing, turn it into a specific open item with an owner rather than silently assuming it is satisfied.
Design sampling to preserve origin
Define how and when parts will be selected, with a rationale tied to the question. Record machine, cavity, shift, material and relevant events. Data quantity, frequency and grouping need technical judgment; this guide does not offer a universal sample size for every study.
Repeated measurements of one part do not represent several produced parts. Keep repeat identities and their purpose. If subgroups are used, document what brings observations together and how groups are separated. Agree on those definitions before presenting the summary to someone who did not collect the data.
| Field | Use during review |
|---|---|
| Time order | Connect values to changes and events |
| Part and cavity | Identify the units being compared |
| Reading and repeat | Separate parts from repeated measurements |
| Documented condition | Define where the conclusion applies |
Review the series before summarizing it
Check units, missing values, duplicates and labels without erasing originals. Record each correction and its reason. If a value looks unusual, investigate its context; removing it simply because it lowers the index prevents the report from representing observations.
The linked NIST material places stability before interpreting capability. It also describes assumptions behind conventional indices, including distribution and variation estimation. Identify the analysis the competent reviewer will perform and retain the result. A list sorted from smallest to largest does not replace the time sequence needed for that review.
Explain the sigma estimate and scope
VYRCO’s Cp/Cpk tool receives limits, mean and sigma. It cannot know how sigma was estimated or whether the data represent the stated conditions. Record estimation method, period, grouping and analysis reference; do not compare indices computed on different bases as though they were equivalent.
The final report must preserve data and limitations, not just a number. Explain which features were evaluated, which conditions were excluded and who reviewed the assumptions. Product or process acceptance depends on agreed requirements; this guide sets no thresholds and grants no automatic approval. Keep a route back to the original records for later questions.
Fictional example: two cavities are not an anonymous list
Lengths from C1 and C2 are collected with time and sample identities. Before calculating a combined index, the team recognizes that observations come from two identifiable groups. It reviews each group in context and documents whether combining them is justified. It does not select whichever grouping produces the highest index. Original readings remain available, and conclusions stay within the studied period.
Mistakes to avoid
- Collecting numbers without preserving the question and population they represent.
- Treating repeated readings of one part as independent produced parts.
- Deleting unusual values without documented investigation.
- Confusing control limits with specification or using sigma without its estimation method.
Checklist before continuing
- Question, scope, feature and revision are defined.
- The measurement system has a documented review or explicit open items.
- Selection preserves time, parts and relevant conditions.
- Corrections and exclusions retain reasons and original data.
- The report identifies assumptions, sigma estimation and limitations.
Sources and review
- NIST/SEMATECH · Production Process Characterization
Planning manufacturing data collection and analysis.
Accessed: 2026-10-10 - NIST/SEMATECH · Assessing Process Stability
Stability and time order as context for interpreting process data.
Accessed: 2026-10-10 - NIST/SEMATECH · What is Process Capability?
Capability index definitions and assumptions; does not set an acceptance criterion for your product.
Accessed: 2026-10-10
Original general guidance by VYRCO based on the listed references. Examples are fictional. Editorial review does not validate your process. To correct an item, identify the article, date and source at info@vyrco.com.