IATF16949 Practical Application (20) | Division of Three-Level Audits: Boundaries, Scoring, and Integration of System Audit, VDA6.3 Process Audit, and Product Audit

A company's annual internal audit has just concluded, and three reports are on the quality manager's desk: the system audit identified 6 nonconformities, all related to "documents not updated in time" and "missing training records"; the process audit covered two lines, with scores above 85 points, concluding that "the process is under control"; the product audit sampled 5 batches, and all dimensions and appearances were qualified. The conclusions of the three reports do not contradict each other, suggesting that the system is running well. However, six months later, during a second-party customer audit, the auditor spent two hours on the same assembly line and issued a serious nonconformity: the tightening torque for a certain safety component was only specified as "torque gun setting value" in the control plan, with no torque monitoring or poka-yoke verification. None of the three internal audit reports mentioned this issue—system audit did not check process parameters, the process audit checklist did not include torque, and the product audit only measured the dimensions of the finished products.

Each of the three audits was conducted, but each only covered a part of the required scope, leaving gaps unaddressed. The issue is not the diligence of the auditors but the lack of clear definitions for three critical aspects: the boundaries of each type of audit, how the annual audit plan demonstrates coverage of all processes and shifts, and how the findings of the three audits can be integrated into a usable system input. This article provides four deliverables that can be directly used: a twelve-dimensional comparison table, an annual audit plan table with frequency grading, VDA6.3 scoring and downgrade criteria, and a product audit sampling and defect coefficient (QKZ) determination table.

1. Boundaries of the Three Audits: Twelve-Dimensional Comparison Table

A typical symptom of mixing the three audits is using a single standard to measure all three: using clause checklists to walk the shop floor or using process scores to review documents. First, define the boundaries in a table and explain each row during internal auditor training. This is more effective than repeatedly explaining "the differences between the three audits."

Three Hard Lines for Boundaries to prevent the three layers of audits from encroaching on each other:

1. System audits do not evaluate the operational details of individual operations but only determine whether the required processes, documents, responsibilities, indicators, and monitoring are complete and effective. When "no torque monitoring" is found, the system audit should conclude "process design output does not include necessary control methods" and then hand it over to the process audit for deeper investigation.
2. Process audits do not primarily judge based on the completeness of documents but focus on actual execution at the site and process capability. Missing documents are just one piece of evidence, not the conclusion.
3. Product audits do not judge the quality of the process but only whether the physical product meets requirements. After a defect is found, the process audit should identify the process cause, while the product audit itself does not perform root cause analysis.

2. Annual Audit Plan: A Table to Prove "Coverage of All Processes and Shifts"

IATF 9.2.2.3 requires the internal audit plan to meet three "musts": it must include system, process, and product audits; it must determine the frequency based on process importance, impact on quality, past performance, and risk; and it must cover all processes and consider shifts. During external audits, this is the most frequently checked—auditors will compare the process list and audit plan to see if any process has not been audited in a year or if night shifts have been audited.

1. Frequency Should Not Be Habitual, Use Four-Factor Grading

The total score ranges from 4 to 12 points, corresponding to the audit frequency:

The Bottom Line Must Not Be Broken: Regardless of the score, each manufacturing process must be covered by a process audit at least once per calendar year. Special processes (heat treatment, electroplating, welding, painting, etc.) must follow CQI series requirements and maintain annual coverage.

2. Essential Fields in the Plan Table

An annual audit plan that can be directly accepted by external auditors must include the following columns, with any missing column being a point of inquiry:

3. Use the "Gap Identification Method" to Self-Check Coverage

Create a matrix with the process list horizontally and the twelve months vertically. Fill each cell with the audit type code (S/P/PD) and then check for gaps row by row and column by column:

• Row Gaps: An entire row for a process is empty → The annual plan is missing an item, directly add it.
• Column Gaps: An entire column for a month is empty → The plan is concentrated at the beginning or end of the year, indicating uneven distribution and typically insufficient depth.
• Shift Gaps: Add the shift as an additional column and fill it again; if only day shifts have codes → The shift coverage requirement of 9.2.2.5 is not met.
• Overload in the Same Month: Three or more rows in a month, with the same auditor → The actual execution will inevitably compress the duration, and the report quality can be predicted.

Printing this matrix and attaching it to the plan is the simplest way to prove coverage. Its role goes beyond just meeting audit requirements: gaps often correspond to "hard-to-audit" processes that no one internally wants to audit.

3. System Audit: Advance with "Clause—Process—Evidence" Three Columns

System audits are often conducted as a reading of clauses. An efficient approach is to create a three-column audit plan: the left column for clauses, the middle column for the processes and documents that support the clauses, and the right column for the evidence to be reviewed.

Input List (must be collected before the system audit starts, otherwise the audit will be superficial): NCs and closure evidence from the last internal and external audits, customer complaints and returns in the last 12 months, customer scorecards, monthly process capability reports, change logs, new project lists, warranty data, supplier performance.

Unified Scoring Criteria to avoid different conclusions from different auditors for the same issue:

The output of the system audit is not just an NC list but must also answer the question: which of the findings are single-point execution issues, and which are planning layer deficiencies? The latter is the input for management review. The judgment method is simple—when the same type of nonconformity appears in three or more processes, or the same process has the same type of nonconformity in two consecutive rounds, it is treated as a systemic defect.

4. Process Audit: Scoring, Downgrading, and "One-Sentence Evidence"

The six process elements of VDA6.3 form the skeleton of the audit: E1 Project Management, E2 Product Development, E3 Process Development, E4 Supplier Management, E5 Process Analysis/Production, E6 Customer Service and Satisfaction. The main battlefield for batch production process audits is E4 to E6, with E5 typically having the highest weight and most likely to expose real issues.

1. Scoring Criteria: What Each Score Range Represents

When scoring, auditors must write a "one-sentence evidence" in the format "what was seen + where + on what day," for example, "2026-09-12 night shift Line 2 torque gun setting value 45N·m, no torque monitoring data, operator unaware of over-torque handling." Scores without one-sentence evidence are not recognized—this rule eliminates 80% of "scoring based on impressions."

2. Compliance Rate and Grade

The element compliance rate is calculated as the weighted average of the scores for each question within the element (EG% = sum of actual scores ÷ sum of possible scores × 100). The overall compliance rate is the weighted average of the element compliance rates based on the number of questions. Grade determination:

3. Downgrade Rules: Do Not Allow Total Scores to Conceal Critical Gaps

A high total score but a collapse in key elements is the most common distortion in process audits. The following scenarios must be downgraded (one grade lower, and cannot be rated A) after the overall grade is calculated, with specific thresholds based on the version used and customer requirements. The criteria provided here can be directly written into the internal audit procedure:

Downgrade rules must be clearly stated in the audit plan and signed off by the audit team leader. Otherwise, once the grade is out, the first reaction of the audited department will be "Why is it still B with a total score of 88?"—the time spent arguing will exceed the time spent on the audit itself.

4. Interface with Product Audit

The on-site evidence from the process audit must include the recent defect list from the product audit. The approach is: functional defects found in the product audit are treated as mandatory items in E5 of the process audit, checking whether the corresponding operation is identified as a risk point, whether the control plan has corresponding control methods, and whether the detection rating in the PFMEA is still valid. This interface ensures that the product audit is not just "measuring dimensions."

5. Product Audit: Sampling, Limit Samples, and Defect Coefficient

The object of the product audit is the physical product, and it is easiest to conduct as "measure a few pieces." To make an effective judgment, three things are needed: a sampling plan that represents the batch, limit samples to judge "boundaries," and a quantifiable metric for horizontal comparison.

1. Sampling: Determine Sample Size First, Then Sampling Method

Sample size is determined based on batch size and inspection level (commonly using general inspection level II sample size):

Three requirements for the sampling method: randomly sample from batches ready for delivery, not from retained samples or pre-selected conforming products; samples from the same batch must cover different production times (e.g., half from the morning shift and half from the night shift); for products with special characteristics, the sample size is doubled based on the above, and the actual values of special characteristics are recorded separately.

2. Limit Samples: Judging Boundaries Cannot Rely on "Master's Eye"

For appearance and fit requirements, there must be limit samples (upper limit, lower limit, boundary sample) that are confirmed by the customer or internally approved and archived. The samples should be labeled with a number, characteristic, approval date, and validity period. During product audits, first check the limit sample cabinet—missing or expired samples mean that appearance judgments lack a comparable baseline, and the "conforming" conclusion of the product audit loses its basis.

3. Defect Coefficient QKZ: Turn "How Many Defects Were Found" into a Comparable Number

Defects are weighted based on their impact, and the weighted sum is divided by the sample size to get the QKZ:

Calculation example: For a product, 20 pieces are sampled, and 1 A-level defect, 2 B-level defects, and 4 C-level defects are found. Weighted sum = 1×10 + 2×3 + 4×1 = 20, QKZ = 20 ÷ 20 = 1.00.

Threshold values should not be arbitrary: use the 80th percentile of the actual QKZ distribution from the previous year as the improvement line and the 95th percentile as the nonconforming line, adjusting annually. This ensures that the threshold is challenging but not so strict that everyone gives up.

4. Product Audit Check Items Are Not Just Dimensions

Complete check items should include at least five groups: characteristic conformity (dimensions, material properties, function), appearance and surface treatment, labeling and traceability (batch number, date code, traceable code readable and verifiable), packaging and protection (including customer-specified packaging norms), and accompanying documents (certificate of conformity, inspection report, packing list). The last three groups are the most frequently overlooked and are also the areas with the highest customer complaints.

6. Internal Auditor Qualifications and Avoidance: Qualification Matrix Is More Important Than Certificates

9.2.2.1 requires internal auditors to have capabilities matching the content being audited and that auditors should not audit their own work. Operationally, this is managed with a qualification matrix:

Three avoidance rules: do not audit your own department, do not audit processes corresponding to documents you have written, and do not audit KPIs you are directly responsible for. When the company size is small and complete avoidance is not possible, supplement with external resources or auditors recommended by customers and note the alternative arrangements in the plan—this is more effective than explaining after the fact.

7. Integration and Linkage of Three-Level Findings

Each of the three audits issues its own report, leading to the same root cause being addressed by three different measures, which may not solve the problem even after all are closed. The integration rules are as follows:

Three linkage paths, written into the procedure document, ensure a true closed loop:

1. Process Audit → System Audit: Quarterly, aggregate the A/B/C grades and elements with compliance rates below 75% from process audits, and use them as mandatory inputs for the next round of system audits. The system audit report should respond to "whether these process gaps indicate systemic planning deficiencies."
2. Product Audit → Process Audit: When A-level defects or QKZ over the line are found in the product audit, conduct a special process audit within 30 days, limited to the operations and shifts corresponding to the defect.
3. System Audit → Next Year's Plan: Systemic defects identified in the system audit directly change the frequency of process audits and the sampling density of product audits in the next year, with the basis noted in the "Change/Risk Notes" column of the plan table.

A quarterly summary table with the following fields is sufficient: finding number, source audit type, process, clause/element, problem description, root cause classification (people/machine/material/method/environment/measurement/planning), measures, responsible person, deadline, verification evidence, whether horizontally deployed. This table is the most practical input for management review.

8. Key Points for Implementation and Common Mistakes

Key Points for Implementation

• First, create a process list and shift matrix, then schedule the annual plan; the plan should fully cover the list, not just visit departments in batches.
• Separate the checklists for the three types of audits: system audits use clauses—process—evidence three columns, process audits use VDA6.3 checklists, and product audits use characteristic lists plus defect grading tables.
• Process audit scoring must include one-sentence evidence, and downgrade rules must be applied before the grade is determined to avoid post-audit disputes.
• Product audits must have approved limit samples and a calculable QKZ; otherwise, the conclusion can only be "looks okay."
• Integrate findings from the three layers quarterly, and use the summary table to judge systemic defects during the annual management review.

Common Mistakes

• Using system audits to replace process audits: clauses are fully covered, but a specific manufacturing process is not audited for the entire year, directly failing 9.2.2.5.
• The annual plan only states "once in the first half and once in the second half," without process and shift dimensions, making coverage unprovable.
• Process audit scores are concentrated at 8 and 10, with compliance rates consistently above 90%, clearly contradicting customer complaints and internal defect rates.
• Judging as A based on total score, ignoring the collapse of individual elements, and still giving A even if a star question is scored 0.
• Product audits only measure finished product dimensions, not checking labeling, packaging, and accompanying documents, leading to rejections after shipment due to labels or documents.
• Using expired, discolored, or worn limit samples, making appearance judgments lose their baseline.
• Auditors auditing their own departments or the processes corresponding to documents they have written, failing to meet avoidance requirements.
• Each of the three layers closing their findings independently, the same root cause leading to three different sets of measures, 100% completion rate, but the problem reoccurs.
• Audit reports only stating "X items found, all rectified," without distinguishing between single-point execution issues and planning layer deficiencies, providing no useful input for management review.
• Completing the annual plan and archiving it without updating for process changes or new mass production projects, leading to a disconnect between the plan and the site.

The value of the three-layer audits does not lie in the number of audits conducted but in the alignment between the layers: system audits prove that all required elements are in place, process audits prove that processes can consistently produce conforming products, and product audits prove that the delivered items are indeed conforming. Missing any one layer can give a false sense of security to the other two.

The three-layer audits should be divided but not isolated, to achieve a true closed loop.

Knowledge code: 2.1.2

Version: v20261009

Author: QTank QTank is dedicated to providing quality management professionals with systematic knowledge, methodologies, and practical tools to continuously enhance the quality capabilities of enterprises.