A factory may reply to a drawing question with a neat spreadsheet and a Cpk value. That result can be useful, but it does not answer every question a buyer needs answered before mass production. A result may cover the wrong drawing revision, a short run after adjustments, one tool cavity, or measurements made with an unclear method. It may also describe a process that was stable for a small sample and then moves after a tool change, material lot change, or shift change.
A China supplier process capability study is most useful when it documents a defined production process for one measurable product characteristic and lets the buyer review the evidence behind the reported index. It should connect the product specification, the measurement method, the production conditions, the raw observations, and the supplier’s follow-up plan. When those pieces are missing, a Cpk value becomes a claim that is difficult to use.
This guide explains how a buyer can request, read, and act on capability evidence from a China factory. It is general sourcing and quality guidance. It does not set a universal acceptance index, prescribe test methods, or replace product-specific engineering, laboratory, regulatory, or legal advice. Set acceptance rules with the people responsible for your product and market.
Start with the decision the study must support
Do not request capability data because it is part of a generic supplier checklist. Request it when a characteristic has a known effect on fit, function, safety, assembly yield, appearance, compliance, or customer returns, and when it can be measured consistently. A capability study can help a buyer decide whether a process is likely to hold a stated tolerance during ordinary production. It cannot prove that every unit in a future shipment will conform.
NIST defines process capability as a comparison of an in-control process output with specification limits through capability indices. Its explanation is a useful warning for buyers: capability discussion starts with process control, not with a spreadsheet formula.1 The buyer should therefore write down the release decision before the factory begins collecting data.
For example, a custom product may use a molded feature that must fit a purchased mating component. The buyer may ask for capability evidence for that feature before authorizing full production. The decision is not “Does the factory have an impressive Cpk?” It is “Does the evidence show that this tool, material, measurement method, and production setup can hold the approved limits, and what will happen if it cannot?”
| Situation | When capability evidence may help | What it does not replace |
|---|---|---|
| New mold, die, fixture, or assembly method | It shows whether a defined critical characteristic has enough margin under stated study conditions. | Tool acceptance, first-article approval, and shipment inspection. |
| Tight fit between two parts | It helps assess variation against the mating-part requirement. | Actual fit testing with the approved mating part. |
| Repeated dimensional customer complaint | It can identify spread, centering, and possible process movement. | Root-cause investigation and verified corrective action. |
| High-volume custom component | It provides a baseline for a repeatable production control plan. | Lot traceability and routine in-process checks. |
| Cosmetic or subjective requirement | It may be less useful unless the characteristic has a reliable numeric method. | Approved appearance samples and visual inspection rules. |
A buyer does not need a study for every drawing dimension. Over-requesting data wastes time and can produce reports that no one reviews. Start with a short list of critical-to-quality characteristics. A characteristic belongs on that list when failure would block assembly, affect stated performance, create a customer-facing defect, or create an unacceptable risk. Record why it was selected, the required limits, and the owner who can interpret the result.
Freeze the baseline before the factory measures parts
The quality of a study cannot exceed the quality of the baseline used to make it. Before measurements begin, give the supplier one controlled package. That package should state the part number, current drawing revision, unit of measure, nominal value where relevant, upper and lower specification limits, test or measurement location, sample condition, and any linked work instruction. If the characteristic depends on a mating part, send or identify the approved version of that part too.
The buyer and supplier should also agree on what counts as normal production for the study. A measurement result taken from parts made during a tool tryout, immediately after manual adjustment, may be relevant to development but may not represent the factory’s standard operating condition. Ask the factory to identify the machine, tool or cavity, material or component lot, operator or shift grouping when material, method, or shift may affect output, and any special setup used during the run.
A drawing tolerance is not the same thing as a control limit. Specification limits express what the product must meet. Control limits come from process data and help a factory decide whether observed variation still resembles the process baseline. NIST describes statistical process control as comparing later data with limits established from earlier data, then investigating points that fall outside those limits.3 Keep the two types of limits separate in meeting notes, reports, and supplier emails.
| Baseline item | Buyer question | Evidence to retain |
|---|---|---|
| Product identity | Which SKU, part number, and drawing revision does this cover? | Approved drawing or specification and controlled revision record. |
| Characteristic | What exact feature is being measured, at which location, and in what unit? | Marked-up drawing, method image, or inspection instruction. |
| Product requirement | What are the target, upper limit, lower limit, and any one-sided limit? | Approved specification, including any buyer-approved deviation. |
| Measurement method | Which gauge, fixture, test setup, resolution, and instructions will be used? | Method sheet, gauge identification, calibration status, and photos where useful. |
| Study conditions | Which machine, tool, cavity, shift, material lot, and production order produced the data? | Study cover sheet and traceable production record. |
| Release rule | Who reviews the data, what evidence is required, and what happens if results fall short? | Signed supplier-quality plan or purchase-order attachment. |
This baseline also prevents a common dispute. A factory may report a favorable result on a less demanding internal control target while the buyer expects the published customer tolerance. Require the report to display the buyer-approved specification limits on the data sheet and chart. If a supplier proposes a different limit, treat it as a technical change request, not a formatting choice.
Request raw data and study context
A useful China supplier process capability study lets a competent reviewer retrace the reported result. Ask for the complete package in a format that can be archived with the purchase order or quality file. A PDF can summarize the study, but the raw observations should also be available in an editable spreadsheet or export. Do not require the supplier to share unrelated proprietary information. Ask for the evidence needed for the specific characteristic and production release.
The request should identify whether the data is grouped by cavity, machine, line, shift, material lot, or time order. Combining unlike sources may hide a difference that matters. For a multi-cavity mold, one overall result can mask a cavity that runs close to a limit. For an assembly line, data from an experienced setup may look different from data after a routine changeover. The buyer should decide which sources need separate reporting based on product risk and how the factory actually makes the item.
Use this request checklist as a starting point.
| Requested item | Why the buyer needs it | Warning sign |
|---|---|---|
| Current drawing and limits shown on the report | Confirms that the calculation uses the approved requirement. | Report lacks revision, unit, or specification limits. |
| Raw measurement table in production order | Allows checking of count, range, missing values, and time sequence. | Only a final Cp or Cpk figure is provided. |
| Sample plan and selection explanation | Shows how many pieces were measured and what the samples represent. | Supplier says samples were “random” without identifying the run or sequence. |
| Production context | Links results to machine, tool, cavities, material, and setup. | Data cannot be traced to a tool, line, or lot. |
| Measurement method | Supports a review of whether the readings describe the characteristic consistently. | No gauge ID, no method, or an unsuitable resolution. |
| Stability plot or control chart | Shows whether mean and variation move over the study sequence. | A histogram is supplied with no time order or control evidence. |
| Capability calculation and assumptions | Lets the buyer see formula inputs and any normality treatment. | Index reported with no mean, standard deviation, or stated assumptions. |
| Nonconformance and action record | Connects an adverse result to containment, correction, and recheck. | Factory provides a passing-looking report but no action for anomalies. |
Ask the supplier to label the study honestly. A short engineering trial, a preproduction run, and an established production study answer different questions. A positive trial result may support more sampling or a pilot run. It should not be presented as long-run evidence unless it includes the agreed conditions and enough representative production time.
Check stability before reading Cp or Cpk
A capability index describes a relationship between variation and the specification. It is not designed to explain a process that is moving. NIST states that a stable process has a constant mean and a constant variance over time, and says that process stability should be demonstrated before capability is discussed.2
For a buyer, this means the data should retain sequence. Review a run chart, scatter plot, or relevant control chart with the measurement order shown. Look for a sudden step after an adjustment, a gradual drift toward one limit, a different band of values from another cavity, or a cluster that matches a material or shift change. These patterns do not automatically mean that the supplier must be rejected. They do mean the team should understand the source before relying on a combined capability result.
NIST notes that graphical methods are usually sufficient for evaluating stability and discusses scatter plots and control charts as practical tools.2 A sourcing team need not conduct a statistical audit on every report. It can ask direct questions: What event corresponds to this shift? Which parts were made before the adjustment? Were all cavities included? Did the factory remove any observations, and why? Was the process restarted after a measurement issue?
A factory should preserve raw values, including values that fail the requirement or trigger investigation. Deleting a reading without an accountable explanation can make a report look better while making it less useful. If a measurement is invalid because of a documented gauge mistake, keep the original entry, identify why it was excluded, and show the corrected measurement process. If the part itself is out of specification, the result belongs in the evidence and should lead to containment or correction.
A buyer-friendly stability review
Read the chart in this order. First, check whether the samples follow the stated production sequence. Second, compare the pattern with material lots, cavity identity, setup events, and shifts. Third, ask what the supplier did with any unusual point. Fourth, confirm whether the post-action data was measured under the same method and product revision. This sequence keeps the conversation anchored to evidence instead of turning it into a debate about one index.
When a factory cannot provide a stability view, do not fill in the gap with confidence. You can still use the available measurements for a preliminary decision, but document the limitation and apply additional controls. Those controls may include a pilot run, increased in-process checks, separated cavity data, an independent inspection, or a hold on volume release until the supplier supplies representative evidence.
Read Cp and Cpk in context
For a two-sided specification, NIST defines Cp as the specification width divided by six process standard deviations. It defines Cpk using the distance from the process mean to the nearest specification limit, divided by three standard deviations.1 Put plainly, Cp describes the spread relative to the available tolerance, while Cpk also reflects whether the average sits too close to one limit.
The difference matters. A process can have a reasonable spread and still be centered poorly. NIST notes that Cpk is less than or equal to Cp because an off-center process loses margin to the nearer limit.1 If a report shows a much lower Cpk than Cp, ask the supplier whether the process needs centering, whether the target is correct, whether one cavity is different, or whether a setup parameter changed.
Avoid treating any index as a portable pass or fail label. Customer requirements, product risk, process maturity, measurement uncertainty, sampling design, distribution shape, and the cost of a failure all affect the appropriate decision. NIST’s formulas also assume normally distributed data for the capability indices it presents.1 A factory should state how it assessed that assumption or what approach it used if the data is not approximately normal. Buyers should send a questionable result to a qualified quality or statistical professional rather than changing the formula in a sourcing meeting.
| What the report shows | Likely interpretation | Buyer follow-up |
|---|---|---|
| Cp and Cpk are close, data is stable, and samples match normal production | The process appears centered within its studied spread and conditions. | Confirm the agreed release rule, store the evidence, and retain routine inspection controls. |
| Cp is higher than Cpk | The process may be off target or closer to one specification limit. | Request a centering analysis, affected-source review, and post-action evidence. |
| Both indices are low or omitted | Variation may consume too much tolerance, or the analysis is incomplete. | Hold the release decision, define containment, and request a corrective-action plan with a repeat study. |
| Good index but chart shows drift, shifts, or segregated clusters | The summary may hide instability or different production sources. | Request separated analysis and review conditions before accepting a combined result. |
| Good result from a small or unclear study | The result has limited decision value. | Request a defined protocol and broader representative evidence. |
| Report uses a different drawing revision or internal limits | The study does not address the approved product baseline. | Reject the report for release purposes and ask for a corrected study. |
Do not confuse a control chart limit with a customer specification. A point inside control limits may still fail a product requirement, and a point inside a product specification may signal a process shift that the factory should investigate. The buyer’s goal is to retain both views: product conformance and process behavior.
Set a release path before the factory sends results
A report is easier to use when the supplier and buyer already know the possible decisions. Write a short release matrix into the quality plan or purchase-order appendix. Tie every decision to evidence, owner, timing, and shipment status. This prevents a rushed call after a supplier finishes a trial run and needs an answer the same day.
| Review outcome | Production status | Required supplier action | Buyer record |
|---|---|---|---|
| Evidence meets the agreed study protocol and release rule | Release under normal agreed controls. | Keep production parameters and routine checks in place. | Approved report, baseline, and date of next review. |
| Evidence is incomplete but product risk permits a limited pilot | Conditional release for the defined lot or quantity. | Add agreed containment, sampling, and deadline for missing evidence. | Written deviation, scope, expiry, and responsible approver. |
| Result indicates centering or variation concern | Do not approve unrestricted volume based on the report. | Contain affected product, identify cause, correct process, and repeat the study. | Corrective-action request and verification evidence. |
| Method, revision, or traceability cannot be confirmed | Pause the capability-based release decision. | Rebuild the study from the approved baseline. | Rejection reason and new data request. |
| Product has regulatory, safety, or high-consequence exposure | Escalate to the designated technical or compliance authority. | Follow product-specific validation and release requirements. | Escalation record and approved disposition. |
Capability evidence works best when it leads to a defined response. A request for supplier corrective action should name the characteristic, affected sources, containment action, suspected cause only if evidence supports it, corrective action, owner, due date, and required effectiveness check. The effectiveness check should not be a new promise. It should show that the supplier used the corrected process under the defined production conditions and then reviewed results against the same approved baseline.
Do not release a shipment merely because a supplier corrected a report format. A better chart is not a better process. The factory must address the condition that made the result unreliable, whether that is a measurement problem, mixed cavity data, unstable processing, incorrect specification, or actual variation.
Keep the study current after release
A process study should have an expiry logic. The buyer does not need to repeat every study on a calendar cycle with no reason, but it should identify changes that may invalidate the original evidence. Common triggers include a new or modified tool, a new cavity, machine relocation, material or component substitution, material supplier change, changed process parameter window, changed inspection method, drawing revision, material corrective action, or a relevant customer complaint.
Link the trigger list to the change-notification process. The supplier should tell the buyer before making a change that affects the agreed characteristic. The buyer can then decide whether a full repeat study, a limited confirmation run, fit test, first-article review, or increased inspection is appropriate. For a low-risk change, a documented review may be enough. For a high-consequence characteristic, engineering or regulatory approval may be needed.
Keep capability evidence in the same document set as the current drawing, first-article approval, control plan, inspection records, nonconformance reports, and change notices. A report that cannot be linked to the product revision or production source has little audit value months later. Traceability also makes it possible to revisit the evidence if a downstream problem appears.
NIST distinguishes process monitoring from lot acceptance sampling. Process monitoring compares new measurements with previously established control limits, while statistical quality control samples completed lots to make an acceptance decision.3 That distinction matters after release. A favorable study can justify confidence in a defined process, but it does not remove the need for the inspection activities the buyer and supplier have agreed for each production lot.
Use this supplier request language
The following wording can be adapted for a purchase-order quality attachment. Fill in the brackets rather than sending it as a vague request for “Cpk data.”
For [part number and revision], provide a process capability study for [defined characteristic and measurement location] before mass-production release. Use the approved specification limits of [lower limit] and [upper limit], in [unit]. Identify the measurement method, gauge or fixture ID, calibration status, study sample count, production sequence, machine, tool and cavity where applicable, material or component lot, and relevant shift or setup conditions. Submit raw measurements in sequence, a stability plot or control chart, the capability calculation with stated assumptions, and a record of any excluded observation with its reason. If the agreed release rule is not met or the data shows an unexplained shift, hold affected product, submit containment and corrective action, and repeat the agreed evidence after correction.
The wording identifies evidence without claiming that one formula fits every product. Add your organization’s approval route and the specific release criteria separately. If the part is medical, electrical safety related, food contact, chemical, children’s, aerospace, or otherwise regulated, have the responsible technical and compliance personnel tailor the package.
Common mistakes that weaken supplier capability evidence
The first mistake is asking for a Cpk number without a product baseline. The second is accepting a report with no raw data or production context. The third is combining cavity, shift, or material data that should be reviewed separately. The fourth is treating a passing study as permission to remove normal inspection and traceability controls. The fifth is using a universal target with no link to product risk or customer requirements.
Another common mistake is sending a study request late in the process, after production has already started. Decide on critical characteristics during development, tooling approval, or preproduction planning. That gives the factory a chance to build the measurement method into its work instructions and control plan. It also gives the buyer time to review weak evidence before cargo is at the port.
A report can be statistically polished and commercially unhelpful. The best report is one that lets both parties make a clear production decision, preserve the evidence, and know what to do after a change.
FAQ
Is a China supplier process capability study required for every product?
No. Request it for measurable characteristics where process variation creates a meaningful commercial, assembly, performance, safety, or compliance risk. For many ordinary products, first-article approval, a clear inspection plan, and supplier process controls may be more useful than a formal capability study.
What is the difference between Cp and Cpk in a supplier report?
For a two-sided specification, Cp compares the allowed tolerance width with the process spread. Cpk also considers the process average’s distance from the nearer specification limit. A lower Cpk than Cp can indicate an off-center process. The result only has meaning with the report’s data, assumptions, and conditions.1
Should I set the same Cpk requirement for every China factory?
No. Set a written release rule for the product and characteristic at issue. Consider customer commitments, consequence of failure, tolerance design, process maturity, available measurement method, and applicable requirements. A qualified engineering or quality professional should help decide when the characteristic is sensitive.
Can a pre-shipment inspection replace a capability study?
No. A pre-shipment inspection samples a completed lot. A capability study examines a defined production process and its variation under stated conditions. Use the appropriate controls for each decision. Neither one replaces the other automatically.3
What should I do if the factory refuses to send raw data?
Ask whether the concern is confidentiality, format, or workload, then narrow the request to the defined characteristic. Offer a protected template or supplier portal if available. If the supplier still cannot provide evidence sufficient for your release decision, document the limitation and choose a different control strategy, such as increased verification, a limited pilot, independent inspection, or escalation.
Make capability evidence a usable release control
A China supplier process capability study has value when it shows more than a number. The buyer should see the approved requirement, a defined measurement method, sequential data from stated production conditions, stability evidence, a transparent calculation, and a response if the process does not meet the agreed release rule. That package makes it easier to decide whether to release, contain, correct, or escalate.
Supplier Ally can help buyers turn critical product requirements into practical supplier evidence requests, review the documents a factory provides, and coordinate follow-up before mass production. The aim is a clear record that supports a production decision, not a ceremonial report that sits unread in an email thread.
References
[1] National Institute of Standards and Technology, “What Is Process Capability?”
[2] National Institute of Standards and Technology, “Assessing Process Stability”
[3] National Institute of Standards and Technology, “What Are Process Control Techniques?”
EN 
ES 