2026.07.09 | Pei-Chi Chiu

AIAG-VDA SPC Manual 2026: Key Changes to Pm/Pmk, Cp/Cpk, Reporting, and Software V&V

For the automotive supply chain, the new manual is about more than changed formulas. The practical questions are how to release equipment after a machine-performance study, how to name capability and performance indices, how to report non-normal data, how to verify and validate analysis software, and whether control charts, OCAP, and the Control Plan are connected on the shop floor.

This article explains the changes that quality, process, supplier-quality, and system-implementation teams should understand first, in language they can use for implementation discussions.

SPC 2026 · Visual guide

PDCA for SPC implementation

01

Plan

Define the study purpose, characteristic, data structure, and decision criteria.

02

Do

Collect time-ordered data under documented and representative conditions.

03

Check

Assess variation, statistical control, model fit, and uncertainty.

04

Act

Standardize effective action, monitor the process, and retain evidence.

Continuous improvement: the result of one study becomes governed input to the next cycle.

1. Not automatically mandatory, but now a common supply-chain language

AIAG and VDA jointly developed the manual to give the North American, European, and global automotive supply chains a more consistent basis for SPC terminology, report formats, and statistical methods. Companies should treat it neither as a new regulation nor merely as a textbook.

A practical first step is to use the manual as an internal gap-assessment baseline. Check whether current SOPs, training, capability-study reports, control-chart settings, software parameters, and customer CSRs align with its concepts. If a major customer or contract formally cites it, translate it into operating requirements.

2. Capability and performance indices: names reflect process status

The manual distinguishes machine performance, preliminary process performance, and production process capability more clearly. An index name is not just a symbol: it indicates whether the data support a stability claim and which release decision the study is intended to support.

Study stage	Common indices	Key judgment
Machine performance study	Pm / Pmk	Used for equipment release. Consecutively produced samples generally assess machine-related variation while operator, material, method, and environmental conditions are held fixed or documented as far as possible.
Preliminary process performance	Pp / Ppk	Used when stability has not been fully demonstrated or the data are insufficient to claim capability. P denotes performance and must not be treated as equivalent to capability.
Production process capability	Cp / Cpk	C-type capability indices are appropriate only after the process has demonstrated statistical control.

A Cpk/Ppk report therefore cannot be judged only by whether its value meets a target. Also check the collection period, sample size, stability assessment, control chart, and whether the index name matches those conditions.

SPC 2026 · Visual guide

Three distinct study stages

01

Machine performance

Evaluate equipment-related variation under defined short-run conditions.

Typical study: 50 consecutive parts · Pm / Pmk

→

02

Preliminary process performance

Evaluate early process output across representative operating conditions.

Performance evidence · Pp / Ppk

→

03

Ongoing process capability

Use capability indices only after statistical control has been demonstrated.

Ongoing monitoring · Cp / Cpk when appropriate

3. Machine performance studies: 50 samples are typical; compensate for smaller samples

The machine-performance section contains an easily missed point: a study normally uses 50 consecutively produced samples. When technical, cost, cycle-time, or sample-availability constraints make 50 impossible, a company cannot simply test fewer parts and retain the same threshold.

SPC 2026 · Visual guide

Sampling across multiple production sources

Station A

Represent each machine, cavity, tool, or stream that can introduce distinct variation.

Station B

Keep time order, traceability, and actual production proportions visible.

Station C

Do not let a large pooled total hide an underrepresented source of risk.

Study design = sufficient total evidence + defensible allocation across relevant sources.

4. Greater emphasis on non-normal distributions, time-dependent models, and reporting

The manual no longer forces every problem into a normal distribution. For geometric tolerances, wear trends, multimodal data, or non-stationary location changes, first understand how the process changes over time, then select a suitable distribution model, control chart, and sampling strategy.

This also changes reporting. The examples emphasize making the source, sample information, distribution model, stability assessment, index calculation, and—when needed—confidence intervals understandable. A single Cpk value cannot communicate sampling uncertainty or process conditions.

SPC 2026 · Visual guide

Capability paths for non-normal data

General geometric method

Estimate relevant distribution quantiles and compare their distance to specification limits.

Report the fitted distribution and assumptions.

z-score method

Convert tail probabilities to equivalent standard-normal z scores for capability interpretation.

Report the transformation and tail model.

First verify process behavior and model fit; do not force non-normal data into a normal model.

SPC 2026 · Visual guide

Traceable handling of suspected outliers

01

Flag

Preserve the original observation.

→

02

Document

Record time, lot, source, and context.

→

03

Investigate

Seek an assignable cause and supporting evidence.

→

04

Decide

Retain or exclude only with traceable justification.

→

05

Verify

Confirm the action and monitor recurrence.

SPC 2026 · Visual guide

What a defensible capability report shows

Time order & control chart

Show signals, limits, subgrouping, and the period studied.

Distribution & model

Model fitProbability plotTail behavior

State assumptions and data treatment.

Indices & uncertainty

Cp / CpkPp / PpkConfidence interval

Interpret values in the study context.

5. Analysis-software V&V: buying software is not the finish line

The manual distinguishes verification from validation for analysis software. Verification confirms that specified requirements and expected results are implemented correctly; validation confirms fitness for intended use under defined operating conditions.

In other words, do not ask only whether the SPC software can calculate Cpk. Also ask:

• Can it retain customer-specific parameter settings and versions?
• Can it identify different calculation methods such as .G and .Z?
• Can it retain sample size, distribution model, outlier treatment, and confidence-interval settings?
• Can it reproduce the statistical conditions used for a historical report?

With manual spreadsheets, the main risk is not whether they can produce a number, but whether parameter transparency, version control, access, traceability, and customer-specific settings can withstand an audit.

SPC 2026 · Visual guide

Software validation and verification

Validation

Does objective evidence show that the software is fit for its intended use in the defined operating context?

Focus: intended use and real workflow

Verification

Does objective evidence show that specified requirements and expected outputs have been implemented correctly?

Focus: requirements and correct results

6. Six checks before implementation

1. Review customer CSRs:Confirm whether major automotive customers cite the AIAG-VDA SPC Manual 2026.
2. Review capability reports:Check whether Cpk/Ppk/Cw/Cwk and Pm/Pmk names align with the stability prerequisites.
3. Review sample-size rules:Confirm sample sizes, sampling frequency, and compensation for insufficient samples in machine-performance and process-capability studies.
4. Review distribution models:Check whether non-normal, time-dependent, multimodal, or geometric-tolerance data are still being forced into a normal model.
5. Review software V&V:Confirm that SPC analysis software is verifiable, configurable, and traceable.
6. Review OCAP:Write control-chart signals, cause investigation, disposition, and escalation paths into shop-floor workflows.

Frequently asked questions

Is the AIAG-VDA SPC Manual 2026 mandatory?

The manual itself is a non-mandatory consensus guideline. It creates specific execution requirements for a supplier when cited by a customer CSR, contract, PPAP, or audit requirement.

Does the new manual mean Cpk can no longer be used?

No. Cpk remains applicable when process stability has been demonstrated. If stability has not been investigated or does not hold, describe performance with Pp/Ppk. Use Cw/Cwk as a diagnostic tool when within-group variation must be analyzed.

Must every SPC report include a 95% confidence interval?

The examples emphasize confidence intervals, but customer and supplier should agree on the actual report format and content. In practice, capability/performance indices, estimated nonconformance rates, and small-sample cases should include a confidence interval or uncertainty statement.

What should a company do first?

Perform a gap assessment covering customer requirements, report templates, sample-size rules, control-chart selection, software V&V, traceability, and OCAP. Waiting for a formal customer demand usually leaves the organization reacting under pressure.