This is Hamamoto from TIMEWELL.
"A single classification takes me half a day or more." "The ministerial ordinance text is so difficult to read that I have to research it from scratch every time." "I'm not confident in my classification results and I lose days waiting for a superior's sign-off." These are complaints I hear constantly from people working in export control.
Export classification is the foundation of export control. Yet in many companies, it has become a hotbed of over-reliance on individual expertise and chronic inefficiency. METI's annual compliance inspections reached 100 cases in fiscal 2023, and a significant number of companies that underwent inspections received remediation guidance specifically on how they were conducting classification.
This article explains how to systematize export classification into five steps, covers the five most common classification errors encountered in practice, and presents a time comparison between manual and AI-assisted approaches.
What You Will Learn from This Article
- A concrete five-step flow for conducting export classification
- A comparison of time and accuracy between manual and AI-assisted classification
- Five real-world classification error patterns and prevention measures
- How to document classification basis and handle audit situations
1. What Export Classification Is
Export classification is the process of determining whether goods you intend to export or technology you intend to provide falls under List 1 of the Export Trade Control Order — the list-controlled items subject to export controls under FEFTA. It is the starting point of FEFTA-based export control. Getting this determination wrong constitutes an unlicensed export and a legal violation.
METI's "Security Trade Control Guidance (Introductory Edition)" (Edition 2.4, January 2025) lays out a four-stage procedure for exporters: export classification, transaction screening, license application, and shipment management. Export classification is the first gate.
Two capabilities are required for classification: understanding the structure of regulatory provisions in the ordinances, and understanding the technical content of the product in question. Personnel who possess both are limited, and the structural problem of work concentrating on specific individuals is one that every company faces.
How to solve export compliance challenges?
Learn about TRAFEED (formerly ZEROCK ExCHECK) features and implementation benefits in our materials.
2. The 5-Step Export Classification Flow
Step 1: Compile the Product's Technical Information
The accuracy of a classification is directly tied to the quality of the technical information going in. Catalog specifications alone are not enough. Gather detailed technical data: design drawings, test reports, specification sheets.
Information to confirm: the product's intended use and function; technical specifications (precision, performance, capacity, frequency range, etc.); component parts and materials; whether software is involved and what it does. Cutting corners here creates rework in later steps. Coordination with the technical department cannot be skipped.
Step 2: Identify Item Numbers That May Apply
Use METI's publicly available "Goods and Technology Matrix Table" to narrow down candidate item numbers based on the product type. CISTEC's "Item-Specific Comparison Tables" are also useful.
| Product type | Item numbers to check first | Key points to check |
|---|---|---|
| Machine tools | 6-2, 6-3 | Positioning accuracy, number of NC axes |
| Electronic components and semiconductors | 9, 10 | Frequency, processing capability |
| Sensors and measuring instruments | 14 | Detection sensitivity, resolution |
| Chemical substances | 3-2, 3-3 | CAS number, composition ratio |
| Software | Column D of the relevant item number | Relationship to the target goods |
| Cryptographic products | 9-1 | Key length, algorithm |
A common oversight at this step is the possibility that a single product falls under multiple item numbers. "I found the item number it seems to match, so that's enough" is not the right approach — identify every potentially applicable item number and work through all of them. Skipping this step is what leads directly to Error 2, which I describe later.
Step 3: Compare the Ministerial Ordinance's Technical Specifications Against Product Specs
For each identified item number, compare the technical parameter thresholds specified in the ministerial ordinance against the product's actual specifications.
The process is straightforward: read the relevant ordinance provision; extract the specified technical parameters; compare those against the product's technical data; determine whether the product exceeds the threshold.
One important point: the principle is to use actual measured data rather than catalog nominal values. Even if the nominal value in the catalog falls just below the threshold, if the actual measured value exceeds it, the product is "controlled."
Step 4: Verify Catch-All Control Applicability
Even when a product is determined to be non-controlled under list controls, the work is not done. A catch-all control check still remains.
Four things to confirm: Is the destination a country or region outside Group A? Is the intended use unrelated to the development of weapons of mass destruction or conventional weapons? Is the end-user not listed on the Foreign User List? Has an Inform (notification) been received from METI?
The October 2025 regulatory revision expanded the scope of catch-all controls to include conventional weapons-related items. The change introduced an approach of designating specific items by HS code, meaning some products that previously required no verification now require it. In my experience, only a minority of small and medium-sized enterprises have fully grasped the impact of this revision.
Step 5: Record the Classification Result and Prepare the Classification Record
Classification results are recorded as an internal document. The items to specify in a classification record are as follows:
| Item | Content | Notes |
|---|---|---|
| Product name | Official name, model number | Include both if the catalog name and internal name differ |
| Technical specifications | Parameter values used for classification | Prioritize measured values |
| Item numbers reviewed | All item numbers checked | For multiple item numbers, record results per item number |
| Classification result | Controlled or non-controlled | Record per item number |
| Basis for classification | Explanation of why that conclusion was reached | Include references to ordinance provisions |
| Date of classification | The date the classification was performed | Indicates status relative to regulatory amendments |
| Classifier | Name and department of the person who classified | Include approver as well |
Since the October 2025 regulatory revision, a more detailed explanation of the basis for non-controlled classifications has been required. It is precisely for non-controlled determinations that the specific reasoning should be thoroughly documented.
3. Comparing Manual and AI-Assisted Classification
Time requirements for export classification vary with product complexity and the classifier's experience level, but here is a summary of trends seen in manufacturing environments.
| Comparison point | Manual classification (traditional) | AI-assisted classification |
|---|---|---|
| Time per item | 2–8 hours (complex products: more than a full day) | A few minutes to 30 minutes |
| Large-volume processing (100 items) | 2–4 weeks | 1–2 business days |
| Recording the classification basis | Created manually by the classifier | Automatically generated as a report |
| Keeping up with regulatory amendments | Manually check for amendment notices and update | System updated within 2 weeks |
| Overlooking multiple item numbers | Prone to errors | Reduced through cross-checking |
| Handling classifier unavailability | Work stalls | System continues to operate |
| Ease of audit response | Difficult to search through paper and Excel | History centrally managed and immediately retrievable |
To be clear: AI assistance is not a replacement for human judgment. What is realistic is a division of labor where AI presents classification results and their basis in advance, and the person responsible reviews that output and makes the final determination. This "let AI do the groundwork, let humans make the decisions" model is, at least for now, the most well-balanced approach in my view.
4. The 5 Most Common Classification Errors and Prevention Measures
Based on CISTEC case studies and METI briefing materials, here are the error patterns actually occurring in practice.
Error 1: Classifying Based Only on Catalog Specifications
Catalogs list the nominal value and standard specifications of a product, but actual performance may exceed those values. It is not unusual for the nominal accuracy to fall just below the regulatory threshold while the actual measured data exceeds it.
The prevention measure is simple: always obtain actual measured data from the technical department for classification. When catalog values and actual measured values diverge, classify using the actual measured values. This is the most fundamental principle, but it gets skipped in busy environments.
Error 2: Failing to Check Multiple Item Numbers
Finding one item number where the product is non-controlled creates a false sense of security, leading to the omission of checking other potentially applicable item numbers. A machine tool's body may be non-controlled, while the control software it runs falls under a different item number as a controlled item.
The approach: use the matrix table to identify all candidate item numbers, turn them into a checklist, and work through them one by one. Methodical as it sounds, this is the most reliable approach.
Error 3: Assuming "Civilian Use" Means Non-Controlled
List control classification is determined by technical specifications, not by intended use. "Our product is for civilian use, so it must be non-controlled" is not a valid argument. Even if the intended use is civilian, a product is controlled if its technical parameters exceed the threshold.
As a side note: in my experience, this error is most common among veteran sales staff. Years of experience can lead to an intuitive judgment of "this product isn't a problem" — which bypasses the classification process entirely. The point that intended use becomes a factor in the catch-all control screening phase, not in the list-control classification phase, needs to be communicated repeatedly across the organization.
Error 4: Reusing Outdated Classification Results Without Updating for Regulatory Amendments
If the regulatory threshold for a product has changed since it was previously classified as non-controlled, the old classification cannot be reused. The Japan Chemical Exporters' and Importers' Association's near-miss case collection also reports risks arising from the reuse of outdated classification results.
Always include the date on classification records, and conduct re-classification of relevant products when regulatory amendments occur. Establishing a schedule for periodic review — at least once per year — is ideal.
Error 5: Data Entry Errors When Registering in the System
The classification itself was conducted correctly, but during entry into the internal system, "controlled" and "non-controlled" were transposed. The consequences are serious, as products that actually require a license could be exported without one.
A double-check after system registration is essential. Building a workflow that separates the registrant and the verifier prevents the vast majority of this type of error.
5. Building Systems to Maintain Classification Quality
Regular Education and Training
Leveraging CISTEC's "Fundamentals of Export Classification Procedures" training and METI's outreach program (free, for small and medium-sized enterprises) to continuously update the skills of classification staff is highly effective. The outreach program dispatches advisors at no cost — it is an opportunity not to pass up.
Building a Classification Database
Archiving past classification results into a database allows reference when classifying similar products, reducing time spent. However, be vigilant about capturing all regulatory amendment updates. As Error 4 illustrates, operating on the assumption that "it was non-controlled before, so it's still non-controlled" is risky.
Preparing and Updating Manuals
Using METI's publicly available "Sample Export Control Procedure Manual" as a reference, build a manual that fits your company's actual workflow. The premise is that the manual will need to be updated every time a regulatory amendment occurs.
6. Streamlining Export Classification with TRAFEED (formerly ZEROCK ExCHECK)
TIMEWELL's export control AI agent "TRAFEED (formerly ZEROCK ExCHECK)" (worried.jp) supports export classification work with the following approach to address the challenges described above.
A multi-LLM consensus method — where multiple AI models perform cross-checks — presents classification results and their basis. The final determination remains with the human, but the accuracy of preliminary screening is improved. Reports explicitly showing the classification basis and referenced regulations are automatically generated and can be used as-is during audits.
The batch processing capability for handling 100 to 200 items at once is well-suited to manufacturers and trading companies dealing with large product volumes. When a regulatory amendment occurs, the system is updated within two weeks, reducing the risk of reusing outdated classification results.
TRAFEED (formerly ZEROCK ExCHECK) complies with METI-designated formats and covers FEFTA and catch-all controls. Implementation can begin as early as the next business day, accessible immediately through a web browser.
Summary
Streamlining export classification is not simply about working faster. It means eliminating over-reliance on individuals and reducing rework while maintaining classification quality. The tools for achieving this are: systematizing the work into a five-step flow, preventing common errors through process design, and leveraging AI-assisted tools.
That said, I don't believe that installing a tool solves everything. The final determination is always made by a human, and ongoing education and training for classification staff remains indispensable. Companies that invest in both tools and people are the ones that ultimately build the strongest frameworks.
For those struggling with the efficiency of export classification, please visit TRAFEED (formerly ZEROCK ExCHECK) (worried.jp). Demos and inquiries are welcome at no charge.
References
- METI, "Security Trade Control Guidance (Introductory Edition), Edition 2.4" (January 2025)
- METI, "Specific Export Classification Cases"
- METI, "Sample Export Control Procedure Manual"
- CISTEC, "Guide to Export Classification"
- CISTEC, "Fundamentals of Export Classification Procedures"
- Japan Chemical Exporters' and Importers' Association, "Near-Miss Case Collection"
- JETRO, "How to Verify Whether Export Goods and Technology Are Controlled"