IATF16949 Practical Application (23) | Annual Full-Size Inspection and Functional Testing: 24 Planning Fields, Sampling Rules, and Nonconformance Handling

One week before the customer process audit, the quality engineer was reviewing full-size reports in the shared drive and found a PPAP dimensional result report from December 2023, with the drawing column labeled as Version A. However, the customer's drawing had been updated to Version C by June 2024—specifically, the tolerance for the mounting hole position was tightened from ±0.5 to ±0.3, and the hole position was moved by 1.5 millimeters. The auditor asked: Where are the full-size inspection records for 2024 and 2025? The response was, "We did a full set during PPAP, and the part hasn't changed since then." The auditor laid out the C-version drawing and pointed to the hole: This characteristic is marked as a key characteristic in your control plan, but there is not a single measurement record for it in the entire company. The nonconformity was recorded under 8.6.2.1—full-size inspection and functional testing were not performed at the frequency specified in the control plan.

Three weeks later, the assembly site reported a 1.2-millimeter interference with the bracket, which was the same hole. Looking back, the issue was not due to the inspector's laziness but rather the company's misunderstanding of "full-size inspection" as a one-time event: once done during PPAP, it was considered complete. IATF requires a periodic process, and the cycle, scope, judgment, and nonconformance handling must all be documented. This article breaks down the annual full-size inspection from the clause into a fillable plan table and a set of judgment rules.

1. The Clause Requires "Completion According to Frequency and Reviewable," Not "One-Time Completion"

The content of 8.6.2.1 is brief, but its implementation involves three key aspects:

The key lies in the first element, "as specified in the control plan." Many companies leave the "full-size" column in their control plan as "per customer requirements" or leave it blank, making the plan table a baseless document that cannot prove the frequency is controlled during audits. The correct approach is to specify the frequency in the control plan (e.g., "every 12 months" or "every 6 months before the customer's annual audit"), and the annual plan table should only be a detailed schedule of this frequency.

Another often overlooked point is that 8.6.2.1 refers to "full-size inspection and functional testing." Measuring only dimensions and not running functional tests is a partial nonconformity. The frequency of functional testing should also be defined in the control plan or test plan, not just mentioned once in the DVP&R.

2. PPAP Dimensional Results Do Not Equal Annual Full-Size Inspection

This is the most frequent conceptual confusion in the field. Both are called "full-size reports," but their purposes, samples, judgments, and validity periods are entirely different:

In a nutshell: PPAP is the "entry ticket," and the annual full-size inspection is the "health check report." The entry ticket cannot be used as a health check report, and the conclusions of the health check report cannot replace the entry ticket.

3. Scope and Frequency: Layering Before Scheduling

Not all parts need to be inspected annually, but they cannot be chosen arbitrarily. The judgment sequence is first to look at hard constraints, then to layer.

Hard Constraints (must be inspected annually, no layering allowed):

1. The customer explicitly requires annual inspection in the CSR, quality agreement, or supplier manual.
2. Parts containing safety characteristics (including regulatory items).
3. Parts listed as key or closely monitored in the customer scorecard or quality agreement.
4. Parts that have experienced batch returns, controlled shipments, or customer complaints related to dimensions in the past 12 months.

Layering Scoring (used when there are no hard constraints):

Total score 14-18 points: every 6 months; 9-13 points: every 12 months; 6-8 points: every 24 months, but must be triggered immediately after a drawing change. The layering rules should be written into the procedure document and re-evaluated annually, otherwise, the layering from three years ago will continue to be used, and the risks will have changed.

There are two practical scheduling constraints:

• Do not concentrate all inspections in the month of the customer's annual audit. The easiest signal for auditors to identify as a fabrication is when 20 reports are executed in the same week. Spread them out quarterly, with 2-4 parts per month.
• Triggered full-size inspections must be scheduled separately, not waiting for the annual window. Design changes (8.3.6), first article after mold repair, and PPAP verification after production line transfer all fall under triggered full-size inspections.

4. Plan Table: 24 Fields, Missing One Is an Audit Gap

Tabularization is the key to whether this task can be managed effectively. A usable annual full-size inspection plan table must have at least the following columns:

Column 20 (calibration expiry) is where many people stumble: even if the report itself is well-prepared, the calibration certificate of the measuring tool expires after the execution date, rendering the entire report invalid. Checking the calibration validity before scheduling is much easier than trying to fix it afterward.

5. Characteristic List: Alignment of Drawing, Control Plan, and Special Characteristics List

The most common failure mode in full-size inspection is not inaccurate measurement but omission. The sources of omission are almost always fixed in four places:

1. Title Block and Technical Requirements: Material, coating thickness, hardness, deburring, cleanliness—these are not dimensions but are characteristics specified in the drawing, and must be covered by functional tests or material reports.
2. Unannotated Tolerances: The drawing notes "unannotated tolerances follow GB/T 1804-m," these dimensions also have tolerances and should be covered in sampling, at least for mating surfaces.
3. Composite Tolerances and Datums: Positional tolerance with MMC modifier, composite positional tolerance—measuring procedures that only calculate one direction are equivalent to not measuring.
4. Parallel Drawings: Customer drawings and internal conversion drawings coexist, with different characteristic numbering systems. A change in one drawing may not be updated in the other.

The solution is to establish a characteristic alignment table that ties together the three numbering systems:

This table must be updated whenever the drawing is revised, and the header should clearly state "based on drawing version + change order number." The fastest way for an auditor to determine if you have done this is to compare the latest drawing with this table—if they do not match, it indicates that the change has not been implemented.

A special note on outsourced parts: The supplier's full-size report can be used as input, but cannot be directly used as your own report. At least three things should be verified upon receipt: whether the drawing version used in the supplier's report matches your current version, whether the sample covers each cavity, and whether the measuring equipment is within the valid calibration period. If any of these three things are missing, you should either retest or request a redo.

6. Sampling Rules: Sample the "Process Coverage," Not the "Good Pieces"

The purpose of full-size inspection is to prove that the process can produce conforming parts under control, so the core of sampling is to cover sources of variation, not to find conforming pieces.

Three hard rules:

• No Picking: Selecting the most "standard" pieces from conforming products for measurement provides no proof and can be escalated to a data integrity issue if detected (by comparing actual measurement records of other pieces in the same batch or significant differences in retest results), which is more serious than dimensional nonconformity.
• Samples Must Be Traceable: Each sample must be traceable to the batch number, production date, shift, equipment/cavity number, and material furnace batch number. This information must be included in the report. Samples without traceability information are equivalent to no samples.
• No "Make-Up": If the cycle is missed and a report is hastily prepared before the audit, the dates will not match the production records, which is a significant flaw. It is better to write "not executed this period, reason + corrective action" than to fabricate dates.

7. Functional Testing: Don't Treat It as a Subordinate to Dimensional Testing

The sources of functional testing items are threefold, and the union of these sources should be taken: verification items in the DVP&R, customer technical specifications/product standards, and internal product standards. This should be compiled into a test item table:

Three common gaps:

1. Test conditions not updated after a customer specification revision (judgment criteria remain old, even if the test is performed diligently, it is invalid evidence).
2. Outsourced tests only provide a "pass" conclusion without original curves and uncertainty.
3. Functional test samples and dimensional test samples are not from the same batch, making it impossible to correlate and analyze issues.

8. Judgment Criteria: Three-Level Conclusions + Handling Rules for Marginal Values

There is a rule that must be written into the procedure document: how to handle marginal values. Measurement uncertainty can consume the tolerance. When the actual measurement falls on the edge of the tolerance band (e.g., tolerance ±0.05, actual measurement out of tolerance by 0.003), directly judging it as nonconforming may lead to false positives, and directly judging it as conforming may lead to false negatives. A feasible rule is:

1. If the out-of-tolerance amount ≤ 2 times the gauge resolution, or ≤ 2 times the measurement uncertainty (U, k=2), judge as "pending."
2. Re-measure the pending piece on different equipment 3 times (not the same machine 3 times, which is not re-measurement but repetition).
3. If re-measurement is still out of tolerance → judge as nonconforming; if re-measurement is conforming → record data and uncertainty from both pieces of equipment, and the conclusion should follow customer requirements or "judgment by uncertainty" (a simple approach: |actual measurement - specification limit| > U is considered out of tolerance).
4. If the same characteristic repeatedly shows marginal values, the issue lies in process capability or the measurement system, not the judgment rule, and should be transferred to PFMEA and control plan revisions.

The threshold for the measurement system itself: the gauges used for full-size inspection should have a Cg/Cgk of at least 1.67 (1.33 is the minimum for general process control), and the resolution should be at least 1/10 of the tolerance. A gauge with insufficient resolution that measures "conforming" is meaningless.

9. Nonconformance Handling: Five Steps and Time Limits

Discovering nonconformities is not difficult, but completing the handling is. Here is a set of executable time limits:

The order of the four investigation paths is crucial. Many people immediately suspect the mold. In practice, mismatched drawing versions and measurement system issues together account for more than half: the report uses an old drawing, the gauge calibration has expired, the CMM program has not been updated with the drawing, the probe is worn. Start with these two, as they are the least costly.

After handling, two areas need to be updated: the detection (D) in the PFMEA (if the full-size inspection was relied upon for detection, it indicates insufficient process control); and the frequency and sample size in the control plan (if the same characteristic is out of tolerance twice in a row, it indicates the original frequency is insufficient).

10. Personnel, Equipment, and Records: The Three Foundations of the Evidence Chain

• Personnel: Full-size measurement personnel must have an authorization document, including training records, practical assessments (using blind samples, repeat measurement of the same piece 10 times, σ not exceeding 1/10 of the tolerance), and review cycles (usually 2 years). CMM programming and operation personnel can be separate, but the program version must be controlled.
• Equipment: Calibration must be within the valid period, gauge capability must meet standards, CMM programs must correspond to the drawing version, and the temperature stabilization time must be sufficient (thin-walled parts should be ≥ 4 hours, environment 20 ± 2℃). Outsourced laboratories must have CNAS or customer-recognized qualifications, and reports must include measurement uncertainty.
• Records: A complete report package consists of six components—cover (part number, drawing version, batch number, execution date, conclusion), balloon diagram, actual measurement value table, functional test record, equipment calibration certificate, and measurement personnel qualification certificate. The retention period follows 7.5.3.2.1, with production parts being "during the part's production period + one calendar year," and longer if required by the customer.

Three aspects of electronic records are often overlooked: the naming rules for scanned documents should correspond to the serial number in the plan table (otherwise, no one will find them five years later), backups should have off-site or cloud copies, and modifications should be traceable (who, when, and what was changed). When an auditor asks, "Has this report been modified?" being able to provide the portal upload timestamp and system logs is more useful than verbal assurances.

11. Pre-Audit Self-Checklist (20 Items)

Among these 20 items, 1, 4, 9, and 12 are the four most common areas for nonconformities and the most time-saving points for self-checks.

12. Implementation Points and Common Mistakes

Implementation Points

1. Frequency should be written into the control plan first, then expanded into the annual plan table, not the other way around.
2. The drawing version is the lifeline of this work; each revision must automatically trigger re-validation, not rely on human memory.
3. Sampling should be designed to cover sources of variation, with samples being traceable, and it is better to have fewer samples than to select pieces.
4. Judgment rules (including handling of marginal values) should be written into the procedure document in advance; every minute of post-incident debate amplifies the risk.
5. Nonconformance handling should have time limits, responsible positions, and update targets; handling without updates is equivalent to not handling.

Common Mistakes

The annual full-size inspection is not a test of measurement technology but of the alignment between documents: drawing version, control plan frequency, characteristic list, sampling records, equipment calibration, and handling closure. Any break in this chain invalidates the entire evidence. By tying these six areas together, the annual inspection transforms from "submitting a PDF each year" into a mechanism that truly mitigates change risks.

Annual Full-Size Inspection: Prove Parts Remain Conforming to Current Drawings

Knowledge code: 2.1.2

Version: v20261009

Author: QTank QTank is dedicated to providing systematic professional knowledge, methodologies, and practical tools for quality management practitioners, helping enterprises continuously improve their quality capabilities.