Self-classification evidence guide: you said EAR99 — here is how to prove it.

1. Why "we said EAR99" is not enough

The most common pain in export-control routine is not "our product is controlled" — it is "we say it is not controlled, and now we must prove it". Customer questionnaires want your basis; auditors want the record from the year in question; customs wants a classification code with a story behind it. EAR99 is a conclusion of "falls under no control-list entry" — and proving a negative takes more process than proving a positive: you must show which entries you examined, which thresholds you compared, and why each was ruled out.

2. What self-classification is, and whether it is legitimate

The Export Administration Regulations allow exporters to determine their products' ECCN or EAR99 status themselves — "self-classification", entirely legitimate and how most companies operate daily. The alternative is requesting an official classification from BIS. The freedom of self-classification comes with responsibility: the consequences of a wrong call sit with the exporter, so "deciding correctly" and "proving it" are inseparable. One reminder: self-classified items shipped under encryption license exceptions carry additional procedural duties, such as the annual self-classification report — classification done does not mean obligations done.

3. The official order of review: classification is a walked process, not a guessed entry

The Commerce Control List's official order of review gives a path you can retell and record:

1. Find candidate entries from product characteristics: by category and technical parameters, list every entry that could plausibly apply — including the inconvenient ones, not just the favorable ones.
2. Compare thresholds entry by entry: put the product's actual parameters (performance, key lengths, frequency, process node…) against each entry's thresholds and notes; every "met / not met" needs a parameter source.
3. Handle exclusion and relief notes: many entries carry exclusion notes or chapter-level carve-outs — whether an exclusion applies deserves the same verbatim reading as the threshold itself.
4. Only when nothing captures the item is it EAR99: after walking every relevant entry with a stated reason for each "not captured", the EAR99 conclusion can stand.

Each step of this walk produces recordable intermediates — the candidate list, the threshold comparison, the exclusion analysis. Those intermediates are your evidence. A conclusion reached by skipping the walk can only be reconstructed from memory when challenged.

4. What a defensible classification record contains

• Inputs and their sources: the value and provenance of every technical parameter used (spec-sheet version, engineering confirmation).
• Regulation version and date: which revision of the text you compared against — regulations change, and a record without a version cannot be reproduced.
• Threshold-by-threshold comparison: met / not met / insufficient for each relevant threshold, with grounds — including exclusion notes.
• Conclusion and scope: the classification, which part numbers it covers, and an explicit statement of what this determination does not cover (for example, no transaction review).
• People and time: who decided, who reviewed, and when — names and timestamps turn notes into evidence.

5. Five common mistakes

1. Judging new shipments against old text: carrying conclusions across amendments without re-review — records must pin the regulation version, and amendments must trigger re-checks.
2. Recording the conclusion but not the basis: a one-line "this item is EAR99" is, under questioning, the same as no record.
3. Treating EAR99 as "no obligations at all": EAR99 is only a classification conclusion; destination, party and end-use screening still apply, as do embargoes and restricted-party lists.
4. Missing procedural duties on encryption items: many encryption items shipped under license exceptions still carry annual self-classification reports or semiannual sales reports — right classification with missed procedure is still a finding.
5. Concluding from "approximately": when a threshold parameter is uncertain, the honest move is to record "insufficient information, pending" — not to pick the value convenient for the conclusion.

6. Disclaimer

This guide is general information, not legal advice; for case-specific classification and obligations, consult your counsel. The process described reflects a general reading of public regulations; always cite the current official text.