IATF 16949:2016 does not mandate automated vision inspection. What it mandates is documented control of nonconforming product, evidence of effective inspection processes, and retention of inspection records in a format that can be audited. Automated vision inspection, configured correctly, satisfies those requirements more completely than manual visual inspection can — but only if you understand what the standard actually asks for.
Section 8.5.4: Preservation — What "100% Inspection" Means in Practice
IATF 16949 Section 8.5.4 requires that the organization identify, handle, package, and protect product during internal processing and delivery to ensure conformity. The specific automotive supplement to ISO 9001 tightens this through customer-specific requirements (CSRs) from OEMs — Ford's Q1, GM's BIQS, Stellantis's SQMS. Each CSR layer adds inspection documentation expectations beyond the base standard.
Under typical automotive CSRs, "100% inspection" means every production unit was individually checked and recorded — not a sample-based statistical inference. A stamping press producing 4,000 brackets per shift that runs 10-piece statistical samples every 30 minutes is not running 100% inspection by this definition. If a SCAR is raised against a bracket from that shift, the question from the customer's SQE is: "Show me the inspection record for part serial number X." There is no record. The response is a disposition letter and a corrective action plan, not a data record.
An automated end-of-line vision system running at the exit of the press captures an image and issues a verdict for every single part. That is 100% inspection with an individual record per part. The record exists. The SQE's question has a direct answer.
Section 8.7: Control of Nonconforming Outputs — The Three Data Requirements
IATF 16949 Section 8.7 specifies what must be documented when nonconforming product is identified. Three things matter for automated inspection systems:
1. Description of the nonconformity. The standard requires that you document what the defect was, not just that a part was rejected. A vision inspection system that outputs "FAIL" with no defect classification satisfies the binary disposition requirement but falls short of full Section 8.7 documentation. The system must classify the defect type: surface scratch, dimensional deviation, missing feature, and so on. Eolvision outputs the defect class and confidence score with every reject verdict. That classification becomes the "description of nonconformity" in your quality record.
2. Actions taken. Section 8.7 requires documented disposition of nonconforming product — rework, scrap, or concession. For end-of-line inspection, this is typically automated: the reject is ejected to a quarantine bin via a PLC-controlled divert gate, triggered by the inspection system's discrete output signal. The action is the ejection. The MES record that captures the rejection event and bin assignment is the documentation of that action.
3. Concession authority (if applicable). If a nonconforming part is passed under a concession, the person granting the concession and the basis for the concession must be documented. A vision system configured with a "Review" verdict tier — where a human inspector makes the final pass/fail call on borderline parts — needs to capture the reviewer's decision and ID in the record. A system that auto-passes all borderline parts without human review creates a Section 8.7 gap if those parts are later found defective.
Section 8.6.2: Layout Inspection and Functional Testing
IATF 16949 Section 8.6.2 requires layout inspection at defined frequencies for production parts — typically annually per customer-specific requirements, or at product change. Layout inspection is a full dimensional check against the drawing. Automated end-of-line vision handles some but not all layout inspection requirements. A 2D vision system measuring edge positions and feature locations can automate the dimensional components of layout inspection if calibrated against traceable reference. 3D structured-light inspection automates profile and flatness measurements.
We're not saying automated vision inspection replaces CMM-based layout inspection for critical dimensions — it does not. CMM measurement of GD&T callouts (true position, angularity, runout) requires probe-based measurement, not image-based. Automated end-of-line vision complements CMM layout inspection: CMM runs periodically on samples, vision runs continuously at 100% coverage. The two data streams together satisfy both Section 8.6.2 layout requirements and the 100% inspection expectation from Section 8.5.4.
Record Retention: What Format, How Long
IATF 16949 Section 7.5 (documented information) combined with automotive CSRs typically requires inspection records to be retained for 15 years for safety-critical parts and 10 years for standard production parts. The format requirement is implicit but important: records must be legible, traceable to the production unit, and retrievable in response to an audit request.
A vision inspection system that stores records in a proprietary binary format with no export capability creates a long-term records management problem. By year 5, the software version that wrote those records may not be running, and the records may not be readable. The practical requirement for IATF compliance is that inspection records be stored in a format that can be exported to CSV, SQL, or standard image file formats, with a clear schema: timestamp, part serial/batch ID, station ID, verdict, defect class, and image file reference.
A Tier 1 brake component supplier in Michigan was audited under an OEM IATF 16949 certification surveillance audit in late 2024. The auditor requested defect records for a specific production lot from 18 months prior. The supplier's automated inspection system stored records in a SQL database with images in a mapped network directory. The supplier produced a filtered export of all inspection events for that lot — 847 records — within 10 minutes of the request. The auditor noted it in the audit record as a documented strength. Suppliers running paper-based inspection logs for the same period typically require 4–8 hours to locate comparable records, if they exist at all.
Preparing for an IATF Audit: Practical Checklist
If your plant is approaching an IATF 16949 certification or surveillance audit, these are the four inspection-related items that most commonly generate audit findings:
- Inspection plan completeness: Does your control plan (CP) reference automated vision inspection as the detection method for end-of-line surface and dimensional checks? If the CP references "visual inspection" and you are running automated vision, update the CP to accurately describe the method. Auditors note discrepancies between CP language and actual inspection method as a potential nonconformance against Section 8.5.1.
- Detection gauge R&R: IATF 16949 Section 7.1.5.1 requires measurement system analysis for monitoring and measurement equipment. For automated vision systems, this means attribute gauge R&R studies — demonstrating that the system correctly classifies known-defect and known-good parts with acceptable repeatability and reproducibility. A gauge R&R study for an attribute decision system (pass/fail) requires running a defined sample of known parts through the system multiple times and calculating the kappa statistic for agreement. Target kappa ≥ 0.75 for automotive production inspection.
- Rejection quarantine traceability: Auditors will ask to walk from a reject verdict in your inspection system to the physical bin where that part currently sits. If your reject divert system does not maintain a record linking the reject event to a specific quarantine bin or lot, that traceability chain is broken.
- Escape response documentation: If an escape reached a customer during the audit period, auditors will review the 8D or corrective action record. The question they ask is whether the inspection system's records showed the escape or missed it, and what the containment action was. Be prepared to show the specific inspection record for the escaped part — what verdict the system issued, and why it was incorrect (if the system issued a false pass).
For the full picture of how Eolvision stores and exports inspection records, and how the system is designed to support IATF 16949 documentation requirements, see the how it works page. For automotive-specific deployment context, the automotive inspection page covers typical control plan integration and inspection station configurations for stamping and sub-assembly lines.