Read process capability with its centering and stability conditions

Keep the requirement separate from observed behavior

NIST distinguishes control limits, used to assess statistical consistency, from specification limits, used to assess the product’s intended function. A stable process can still fail its specification. Do not rename a process-derived boundary as an approved product tolerance.

For a fictional finished card, obtain the approved dimension from the responsible product owner and record its unit, measurement location and applicable version. A designer’s nominal layout size is not automatically the agreed manufacturing requirement. Resolve whether the record refers to the artwork, a printed sheet or the final cut item before comparing any measurements.

Ask how the measurements were obtained and which physical units they describe. Retain time and production-run identity rather than accepting an unlabeled column of convenient readings. Have the qualified owner address measurement suitability and the relevant process review. The coordinator’s job is to connect the evidence, not to substitute a spreadsheet summary for an unexplained inspection method.

Preserve the provenance of both the approved requirement and the analytical inputs. If a report was prepared before a material or finishing change, flag that mismatch. Do not repair it by editing the tolerance until the old chart looks reassuring. A requirement change needs its own authorized decision and should not be disguised as data cleaning.

Calculate what the two indices actually compare

NIST defines Cp = (USL − LSL)/(6σ) and Cpk as the smaller of (USL − μ)/(3σ) and (μ − LSL)/(3σ). These normal-process capability interpretations require an in-control process. Sample estimates replace population parameters and introduce estimation uncertainty.

Use stipulated population values, not measured estimates: LSL = 99.5 mm, USL = 100.5 mm and σ = 0.1 mm. At μ = 100 mm, Cp = Cpk = 5/3. At μ = 100.4 mm, Cp remains 5/3 while Cpk becomes 1/3.

Keep both input sheets and both outputs visible. The executed fixture uses exact rational arithmetic, then separately checks the same formulas in decimal arithmetic. It does not generate observations or fit a process model. The numerical contrast is a diagnostic teaching device, not evidence about the cards supplied by any real printer.

Have the analyst identify which quantity appears in the actual report and why it is appropriate. A generic column heading such as capability is insufficient when different calculations can produce different values from the same supplied information. Preserve the symbol, input definitions, units and estimation method rather than copying only the more favorable number into a client-facing summary.

Separate arithmetic agreement from evidence of suitability

The unchanged Cp does not detect this displacement; Cpk reflects the nearer specification boundary. The fixture assumes the stated population conditions. It does not establish stability, normality, estimation precision or a universal pass threshold for an actual process.

Request the relevant analytical evidence instead of trying to infer it from a clean-looking result table. Ask the responsible specialist what was examined, which production period it covers and what unresolved conditions limit the interpretation. A short numerical exercise cannot replace that review, and a coordinator should not certify it merely because both calculation routes agree.

Keep the approved reporting purpose in view. A supplier-comparison slide, a problem investigation and a decision about released goods may require different evidence and authority. Record the intended use before attaching a conclusion. The worksheet can reveal a missing explanation without deciding that a supplier is incompetent or that a delivered item must be rejected.

If the actual report uses a different method, do not silently coerce it into this example. Ask its author to explain the definition and conditions. Preserve a genuine disagreement for qualified review rather than averaging incomparable scores. The aim is a traceable interpretation, not a single dashboard number obtained at any cost.

Write a bounded production handoff

Assemble the approved dimensional requirement, the input definitions, the two checked calculations and the outstanding evidence request. Mark the entire exercise as fictional wherever it is circulated for training. Keep it separate from records used to release goods so that a well-presented example cannot be mistaken for an inspection already performed.

Give the operational owner a concrete question to answer. For example, request the scope and justification of the reported process analysis before using its result in a production claim. Avoid issuing a vague demand for better quality when the actual problem is an unidentified calculation or missing evidence about its applicability.

Once the appropriate evidence exists, have the responsible owner review the proposed wording against that evidence and the agreed use. Retain any unresolved limitations beside the conclusion. A completed interpretation sheet does not change a purchase agreement, adjust a cutting machine, approve an invoice or authorize a public allegation about a supplier.

Save the fixture as a review aid for later reports. When a new summary arrives, reconstruct its meaning rather than assuming that a familiar score label guarantees a familiar method. The useful output is an explanation another reviewer can follow, with a clear boundary between verified arithmetic and the wider production decision.

Make the reasoning concrete.

A fictional print coordinator receives two summaries and prefers the one that makes an upcoming delivery easier to approve. The underlying production evidence has not been supplied.

The coordinator preserves the unresolved questions and requests the relevant report from its author. No equipment setting, product tolerance or acceptance status changes.

Take the questions into your next brief.

Use these prompts in your own notes. Nothing is submitted from this page.

1. Requirement

   Which approved physical dimension, unit and version govern the comparison?

2. Inputs

   What measured or stipulated quantities enter the reported calculation?

3. Contrast

   Can the two fictional alternatives be independently reconstructed?

4. Evidence

   Which conditions has a qualified analyst actually examined?

5. Authority

   Who can use the result for the stated production decision?

Watch for the wrong conclusion.

• Editing a requirement to make an existing result appear favorable.
• Copying an unlabeled summary number without its definition.
• Presenting a training calculation as an inspection already completed.

Sources & scope

Original fictional dimensions and review procedure. Named sources support the statistical distinctions. Only local arithmetic was executed; no physical process, inspection result, control chart or capability estimate was fitted or certified.

• NIST: What is Process Capability? →Primary statistical handbook · Supports: Supports the capability definitions and their conditions.
• NIST: What are Variables Control Charts? →Primary statistical handbook · Supports: Distinguishes control limits from specification limits.