Is Cmk 1.67 a rule or a convention?

It is a published requirement in Bosch Booklet No. 9, edition 11.2019, which is a real Tier-1 rule set applied to Bosch's suppliers, and it is widely used as a default across German automotive. It is not a universal legal threshold. Your customer's own document decides, and some customers are stricter: Mercedes-Benz requires Ppk or Pmk of at least 2.0 for measurable DS and DZ characteristics.

How many parts does a machine capability study need?

Bosch specifies at least 50 parts, with 100 recommended, from one continuous run. Volvo requires at least 30 pieces for Cpk studies on special characteristics taken from the significant production run. Ficosa requires at least 125 units with at least two per independent cavity or tool. The sample rule travels with the customer, and so does the population the sample must be drawn from.

Does VDA 5 set Cmk limits?

No. VDA Volume 5 covers measurement and inspection process suitability, currently in its 3rd edition of July 2021. The attribution of Cmk limit values to VDA 5 appears in at least one widely read vendor article and is wrong. Machine and process capability sit with ISO 22514 and VDA Volume 4, and the openly published acceptance values used across German automotive come from Bosch Booklet No. 9.

Has Cmk been replaced by Pmk?

Not as of 22 September 2026, as far as we could verify. A joint AIAG and VDA SPC manual is in draft as a yellow volume, with a public review window that closed on 3 May 2026, and that draft does replace Cm and Cmk with Pm and Pmk. We did not confirm whether a final version has been published. Until it is, the existing customer requirement applies.

What do we have to keep if the customer never asks for the raw run?

Keep the run window record anyway: timestamps, machine, tool, material lot, operator, parameter set, and the machine's own stoppage and tool-change record for the same window. Most customers ask for the index. The ones who challenge it ask how the study was run, and that question cannot be answered retrospectively.

Home

Compliance · Machine and process capability · status 22 September 2026

The machine capability study is a conditions problem, not a maths problem

For quality engineers and plant managers at automotive and industrial suppliers. What the published acceptance values are, whose rules they come from, and what records you have to be able to produce to show the study ran under the conditions it claims.

In shortAnyone can compute Cmk. The index means something only if the run behind it was uninterrupted, on one material batch, with one operator and no tool change, over 50 to 100 consecutive parts. Bosch Booklet No. 9 sets Cm and Cmk at 1.67 and the follow-on Cp/Cpk and Pp/Ppk at 1.33. The evidence that decides the study is the machine record of the run window.

What this is based onBosch Booklet No. 9, "Machine and Process Capability", edition 11.2019, is published openly and is the clearest real rule set available. Customer requirements from Mercedes-Benz, Volvo Group, Stellantis and Ficosa are read from their own published documents. Nothing is reproduced, only paraphrased. ISO 22514, the customer document and your licensed standard prevail over this page.

The acceptance values, and where each one comes from

There is no single European number. There is a widely used default, and then customer-specific values that override it. Both matter, because a supplier that applies the default to a characteristic the customer has classified differently has done the study for nothing.

Published capability acceptance values and their sample rules
Study	Sample rule	Acceptance	Source
Machine capability, Cm and Cmk	At least 50 parts, 100 recommended, from one continuous run under conditions that suppress non-machine influences	Cm and Cmk at least 1.67	Bosch Booklet No. 9, edition 11.2019
Long-term process capability, Cp and Cpk	At least 125 measurements, typically 25 samples of 5, over a sufficiently long period; roughly 20 production days as a guide	Cp and Cpk at least 1.33	Bosch Booklet No. 9, edition 11.2019
Long-term process performance, Pp and Ppk	Same 125-measurement rule	Pp and Ppk at least 1.33	Bosch Booklet No. 9, edition 11.2019
Measurable DS/DZ characteristics	In deviation from the industry standard	Ppk or Pmk at least 2.0; for stable processes Cpk at least 1.67; 100% inspection until capability is verified	Mercedes-Benz Special Terms, MBST 13/19 section 5, edition 2019-11
Special characteristics	At least 30 pieces from the significant production run	Cpk study required	Volvo Group CSR v2, clause 8.3.4.4, 28 February 2025
Special-characteristic ("CO") parts	At least 125 units, at least 2 per independent cavity or tool	Ppk above 1.67 at initial approval, Cpk above 1.33 thereafter	Ficosa Supplier Quality Manual, edition 6, 20 February 2026
Characteristics on the Product/Process Information Sheet	Inspection frequencies aligned to the targets	Cp and Cpk targets carried on the sheet itself	Stellantis CSR v1, clause 8.2.3.1.2, June 2025

Ficosa is a Tier 1 setting requirements on its own suppliers, not an OEM. It is here because it shows how OEM requirements cascade, not as evidence of what an OEM demands.

Two things follow. First, the Mercedes rule has an operational consequence, not just a numerical one: until capability is verified for a measurable DS or DZ characteristic, those characteristics are inspected 100%. That makes the verification date itself a record, and every part made before it needs the 100% inspection record behind it. Second, the Volvo and Ficosa rules define which parts, not just how many: parts from the identified significant production run, and at least two per cavity. A sample of the right size drawn from the wrong population fails.

What "uninterrupted, single batch, single operator, no tool change" means in records

The machine study deliberately excludes the disturbances that series production will add. That is the whole point of it: it isolates the machine. Every source that describes the study specifies the conditions, and the conditions are what you have to be able to evidence.

Study condition to the record that proves it
Condition	What has to be shown	How it usually fails
Uninterrupted run	A machine timeline for the study window with no stoppage events between the first and last part	A short stop nobody logged. The measurement file is continuous because measurements are continuous; the run was not.
Consecutive parts	Part sequence tied to the run, in production order	Parts measured later in whatever order they came off the tray, so sequence information is lost and any trend is unreadable
One material batch	The material lot consumed during the window, from the material record	Two lots crossed mid-run and nobody recorded the change
One operator	Who was logged on for the window	Shift change inside the window
No tool change	Tool or insert change events for the asset, showing none in the window	A worn insert swapped during the run because it was the sensible thing to do at the time
Stated time window	Start and end timestamps for the run, matching the measurement records	A study dated to the day, not the window, so it cannot be checked against anything

The practical pointA spreadsheet of 50 measurements proves that 50 parts were measured. It does not prove that the run was uninterrupted, that the batch did not change or that no tool was swapped. A continuous machine record for the same window does. If the conditions cannot be evidenced, the index is a number without a claim attached.

The obligation does not stop at release

Bosch requires the process to remain permanently stable after release, monitored with control charts, with the reactions defined in the control plan and documented when they happen. In audit terms that produces three continuing records:

The control-chart record itself, retained, with every rule violation visible rather than smoothed out of a monthly summary.
The reaction to each violation: what was stopped, when, and what happened to the parts made since the previous good check. Auditors reviewing SPC charts and capability values expect a reaction plan where capability falls below requirement, and expect to see it used.
The actual check frequency against the planned frequency, per characteristic. Stellantis aligns inspection frequency to the Cp and Cpk targets on the Product/Process Information Sheet, which makes the achieved frequency an auditable number in its own right.

The recurring weakness here is not the statistics. It is that the measurement record and the production record live in different systems, so nobody can say which machine made the part that produced a given reading, in which time window. Where a capability value cannot be resolved to a characteristic, an operation, a machine and a period, it cannot be defended. For the maintenance side of the same argument, see our page on IATF 16949 clause 8.5.1.5 and TPM objectives.

Free tool

OEM Capacity & Audit Evidence Pack

Capacity declaration with demonstrated rate, a run-conditions log for capability studies, an audit evidence index, and an EnPI and baseline sheet. Excel.

Get the evidence pack →

Correction: VDA 5 is not about machine capability

This mis-citation circulates widely, including in vendor articles that otherwise get the numbers right. VDA Volume 5 is about measurement and inspection process suitability — Prüfprozesseignung — currently in its 3rd edition, July 2021, with a practical guide published in September 2022. It tells you whether your measuring equipment and measurement process are good enough to judge the characteristic. It does not set Cmk limits.

The documents that do sit behind machine and process capability are ISO 22514, specifically ISO 22514-3 for machine performance studies on measured data from discrete parts, and VDA Volume 4 for VDA's process-assurance methods. Bosch Booklet No. 9 cites ISO 22514-2 as its primary standard alongside ISO 22514-1, -3, -5 and -6, ISO 9001, VDA 4 and the AIAG SPC and PPAP manuals. If someone hands you a capability requirement attributed to VDA 5, ask which document it actually came from before you design a study around it.

The distinction is not pedantic. A measurement process that is not suitable produces capability indices that are wrong in an unknown direction, and measurement systems analysis is consistently among the most frequent IATF 16949 nonconformances. Getting VDA 5 right is a prerequisite for the capability study, not a substitute for it.

The terminology change to watch, and its current status

A joint AIAG and VDA SPC manual is in draft as a yellow volume for public review, with a feedback window that ran to 3 May 2026. Its stated intent is to reconcile the long-standing AIAG and VDA differences and to align with ISO 22514, ISO 3534 and ISO 7870.

It replaces Cm and Cmk with Pm and Pmk, on the reasoning that a short-term machine trial says nothing about stability, so the result should be reported as performance rather than capability.
It sets out a sequence: machine performance at qualification, preliminary process performance at qualification, then continuous monitoring with Cp/Cpk and Pp/Ppk after launch.
It moves towards control charts with adjustable significance levels rather than fixed limits that depend on sample size.

Status, plainlyThis is a draft in review, not a published requirement. We did not confirm whether a final version has issued since the May 2026 feedback deadline. Do not rewrite your capability procedure on the strength of it, and do not let a supplier or a customer tell you Cmk has been abolished. Check the published volume before changing anything.

What to have ready before the study

Confirm the characteristic classification with the customerWhether a characteristic is DS, DZ, special or ordinary decides which acceptance value applies. Mercedes at 2.0 and a default at 1.67 are different studies.
Confirm the measurement process is suitable firstVDA 5 or the equivalent MSA work comes before the capability study, not after it.
Fix and record the run windowStart and end timestamps, machine, tool, material lot, operator, parameter set. Write them down before the run, not from memory afterwards.
Keep the parts in sequenceLabel in production order. Without sequence there is no trend analysis and no way to relate a drift to anything that happened on the machine.
Pull the machine record for the same windowStoppages, tool changes, operator logons, parameter changes. This is the document that turns the index into evidence.
Plan the follow-on study nowThe 125-measurement process study over roughly 20 production days is a separate exercise with a separate sampling scheme. Decide the scheme before series production starts, not when the customer asks.

Questions

Is Cmk 1.67 a rule or a convention?: It is a published requirement in Bosch Booklet No. 9, edition 11.2019, which is a real Tier-1 rule set applied to Bosch's suppliers, and it is widely used as a default across German automotive. It is not a universal legal threshold. Your customer's own document decides, and some customers are stricter: Mercedes-Benz requires Ppk or Pmk of at least 2.0 for measurable DS and DZ characteristics.
How many parts does a machine capability study need?: Bosch specifies at least 50 parts, with 100 recommended, from one continuous run. Volvo requires at least 30 pieces for Cpk studies on special characteristics taken from the significant production run. Ficosa requires at least 125 units with at least two per independent cavity or tool. The sample rule travels with the customer, and so does the population the sample must be drawn from.
Does VDA 5 set Cmk limits?: No. VDA Volume 5 covers measurement and inspection process suitability, currently in its 3rd edition of July 2021. The attribution of Cmk limit values to VDA 5 appears in at least one widely read vendor article and is wrong. Machine and process capability sit with ISO 22514 and VDA Volume 4, and the openly published acceptance values used across German automotive come from Bosch Booklet No. 9.
Has Cmk been replaced by Pmk?: Not as of 22 September 2026, as far as we could verify. A joint AIAG and VDA SPC manual is in draft as a yellow volume, with a public review window that closed on 3 May 2026, and that draft does replace Cm and Cmk with Pm and Pmk. We did not confirm whether a final version has been published. Until it is, the existing customer requirement applies.
What do we have to keep if the customer never asks for the raw run?: Keep the run window record anyway: timestamps, machine, tool, material lot, operator, parameter set, and the machine's own stoppage and tool-change record for the same window. Most customers ask for the index. The ones who challenge it ask how the study was run, and that question cannot be answered retrospectively.

Sources

Published by TEEPTRAK SAS, which makes production-monitoring and OEE software. Every figure is sourced on the page. Funding rules, standards and reporting duties change: check the official documents before you budget or commit.