How importers argue
A survey of what requesters actually wrote in each rationale field, across 1,111 of the 1,111 machinery exclusion requests (100%). Every rule is published so it can be argued with, and every request that matched no rule is counted.
This docket has no outcomes. Nothing below can be related to a grant rate, and nothing below says which argument works. It describes how importers write, in a process that closed in March 2025 without a published decision.
Structured answers — no inference required
These fields are yes/no or short-form on the request form. They are counted exactly as filed.
Subject to an AD/CVD order
| Answer as filed | Requests |
|---|---|
| No | 1,104 |
| Yes | 7 |
Will be used for domestic manufacturing
| Answer as filed | Requests |
|---|---|
| Yes | 1,083 |
| No | 28 |
Purchased from US or third country in past 5 years
| Answer as filed | Requests |
|---|---|
| No | 342 |
| No. | 246 |
| Hitachi's heat exchangers are custom designed due to their unique incorporation into compressed air systems. Due to the | 124 |
| No, for the reasons articulated above. | 31 |
| Welly has not purchased any manufacturing equipment from the US or a third country in the last 5 years. | 31 |
Filed by a third party (law firm, association, broker)
| Answer as filed | Requests |
|---|---|
| No | 758 |
| Yes | 353 |
Free-text rationale fields — machine-coded survey
A request can match more than one code, so percentages do not sum to 100. The unclassified row is the share that matched no rule at all — published because a taxonomy that misses half its corpus should say so.
US availability
n = 1,111 · median answer 329 characters (10th–90th percentile 67–1,125)
| Code | What the rule looks for | Requests | Share |
|---|---|---|---|
| AVAIL.NOT_AVAILABLE | States the equipment is not available or not made domestically | 381 | 34.3% |
| AVAIL.LEADTIME | Cites lead time or delivery delay | 298 | 26.8% |
| AVAIL.CAPACITY_OR_SPEC | Cites capacity, capability or specification limits | 292 | 26.3% |
| AVAIL.PRICE | Cites price or cost | 209 | 18.8% |
| AVAIL.SOME_US_EXISTS | Concedes some US source exists | 147 | 13.2% |
| AVAIL.NONE_US | States there is no US manufacturer or source | 108 | 9.7% |
| AVAIL.NO_COMPARABLE | States no comparable or equivalent equipment exists | 76 | 6.8% |
| UNCLASSIFIED | Matched none of the rules above | 189 | 17% |
Third-country availability
n = 1,111 · median answer 326 characters (10th–90th percentile 10–884)
| Code | What the rule looks for | Requests | Share |
|---|---|---|---|
| THIRD.QUALITY_OR_SPEC | Cites quality, specification or capability difference | 526 | 47.3% |
| THIRD.EXISTS | Concedes a non-China source exists, or names one | 471 | 42.4% |
| THIRD.PRICE_OR_LEADTIME | Concedes a source but cites price or lead time | 304 | 27.4% |
| THIRD.SEARCHED_NONE_FOUND | Searched third countries and found nothing suitable | 82 | 7.4% |
| THIRD.NONE | States plainly there is no third-country source | 45 | 4.1% |
| THIRD.UNAWARE | Answers on knowledge or belief | 43 | 3.9% |
| THIRD.CHINA_ONLY | Argues China is the only or dominant source | 18 | 1.6% |
| UNCLASSIFIED | Matched none of the rules above | 302 | 27.2% |
Sourcing attempts
n = 1,111 · median answer 360 characters (10th–90th percentile 112–691)
| Code | What the rule looks for | Requests | Share |
|---|---|---|---|
| SOURCING.EFFORT | Describes an effort, search or approach to an alternative supplier | 615 | 55.4% |
| SOURCING.REJECTED_FOR_REASON | Found an alternative and rejected it for a stated reason | 196 | 17.6% |
| SOURCING.NOT_IDENTIFIED | States no alternative source could be identified | 57 | 5.1% |
| SOURCING.QUOTE_OBTAINED | Obtained or discusses a quote or bid | 51 | 4.6% |
| SOURCING.NAMED_ALTERNATIVE | Names specific alternative suppliers considered | 33 | 3% |
| SOURCING.BCI_DEFERRED | ⚠ Defers the answer to the confidential version — the public record carries no substance here | 4 | 0.4% |
| SOURCING.NONE_STATED | States no attempt, or answers N/A | 2 | 0.2% |
| UNCLASSIFIED | Matched none of the rules above | 389 | 35% |
Supports domestic manufacturing
n = 1,111 · median answer 649 characters (10th–90th percentile 229–1,759)
| Code | What the rule looks for | Requests | Share |
|---|---|---|---|
| DOMESTIC.JOBS | Cites employment | 598 | 53.8% |
| DOMESTIC.INVESTMENT | Cites investment or capital outlay | 405 | 36.5% |
| DOMESTIC.SUPPLY_CHAIN | Cites supply-chain resilience or reshoring | 295 | 26.6% |
| DOMESTIC.NEW_CAPACITY | Cites new or expanded domestic capacity | 72 | 6.5% |
| DOMESTIC.IMPORT_SUBSTITUTION | Cites replacing imports with domestic production | 15 | 1.4% |
| UNCLASSIFIED | Matched none of the rules above | 289 | 26% |
Strategic importance / MIC2025
n = 1,111 · median answer 333 characters (10th–90th percentile 11–967)
| Code | What the rule looks for | Requests | Share |
|---|---|---|---|
| STRAT.DENIES | Denies strategic importance or MIC2025 relation | 820 | 73.8% |
| STRAT.UNAWARE | Answers on knowledge/belief rather than substance | 82 | 7.4% |
| STRAT.DISTINGUISHES | Acknowledges the sector is listed, then distinguishes the specific product | 29 | 2.6% |
| STRAT.CONCEDES | Concedes strategic importance | 11 | 1% |
| UNCLASSIFIED | Matched none of the rules above | 256 | 23% |
A rule that was wrong, and how
STRAT.CONCEDES — The first version fired on 56 requests, of which 47 also matched STRAT.DENIES — an 84% false-positive rate. It matched the phrase "is strategically important or related to…" inside sentences that negate it earlier, e.g. "We are not aware that this manufacturing equipment is strategically important or related to ‘Made in China 2025’". It would have reported ~23% of requesters conceding strategic importance.
A second version using JavaScript lookbehind was worse (35.8%), because lookbehind cannot span the preceding clause. The third and current version is compositional: the affirmative phrase must match AND neither a denial nor a knowledge-qualifier may appear anywhere in the answer. That yields 2.5%, consistent with 81.5% denying outright.
Recorded rather than quietly corrected. A keyword rule over legal prose fails first on negation, and a rule whose count is implausible next to its logical opposite is the signal to look for it.
What this survey is not
- This docket has no published outcomes. No code below can be related to a grant rate, and none is.
- Codes are assigned by the literal regular expressions printed with each row. A request can match several codes, so percentages do not sum to 100.
- Requests matching no rule for a field are counted as UNCLASSIFIED and published as such. Nothing is assigned a code by elimination.
- This is a machine-coded survey intended to test the pass-1 taxonomy against real language, not an authoritative classification of any individual request.