Duty drawback methodology

Written for a customs-broker audience and for CBP audit defense.

Legal basis: 19 U.S.C. § 1313 (Drawback and refunds); 19 C.F.R. Part 190 (Modernized Drawback, the TFTEA regulation effective for claims filed on or after Dec 18, 2018); 19 C.F.R. Part 191 (legacy, retained only for pre-TFTEA claims). Trade Facilitation and Trade Enforcement Act of 2015 (TFTEA), Pub. L. 114-125.

Last reviewed: 2026-07-17. Reflects current tariff law as of this date, including the Feb 20, 2026 Supreme Court ruling on IEEPA (see §4.3).

---

0. How to read this document

Drawback is the refund of up to 99% of duties, taxes, and fees paid on imported merchandise when that merchandise (or a qualifying substitute) is subsequently exported or destroyed. It is the only part of the customs workflow that returns cash to the importer rather than costing them money — which is exactly why it is the highest-intent module for a broker and why the methodology has to be defensible line-by-line.

This document is the complete methodology a broker uses to take a book of cleared entries and turn it into filed, audit-ready drawback claims. It is organized as the actual decision path:

1. Which drawback type applies (§2) — the full statutory taxonomy.
2. Is the claim eligible at all (§3) — importer, time window, provisions.
3. Which duties are recoverable (§4) — the part most methodologies get wrong today.
4. How the import is matched to the export (§5) — direct ID vs. substitution.
5. How the recoverable amount is calculated (§6) — per-unit averaging, lesser-of, manufacturing methods.
6. How the claim is proven and filed (§7–§8) — records, proof of export, ACE type-47, privileges.
7. How the claim is defended (§9) — audit posture.

Each section states the rule, the authority, and — set off in a Commers: note — exactly what the platform automates today vs. what remains a broker judgment call. We never represent a step as automated when a licensed broker must still sign it. See Drawback Advisor for the product/agent spec and the code map.

---

1. The governing principles (apply to every claim)

These five rules are invariant across all drawback types. Every calculation in §6 is a specialization of them.

Commers: Principle 1 is enforced deterministically — the recoverable formula caps at 0.99× the eligible base. Principle 2 (5-year window) runs off the shipment's CBP entry date (the date of importation, falling back to the record-creation date only when the entry date is absent) — not the row-ingestion timestamp — so the window filter, the stamped import date, and the "expiring < 180 days" forfeiture flag all track the true drawback clock. Principle 4 is enforced across runs and matchers: within a run each export and import line is consumed once (quantity-aware after the per-unit cap), and across runs each matcher's reset clears only its own claim types while both matchers exclude any import already anchoring an active claim of the other type — so an import can never be designated as both unused and manufacturing drawback. Principles 3 and 5 are broker/importer obligations the platform supports (it generates the ACE-ready entry and preserves the full record trail) but does not discharge on the broker's behalf.

---

2. The complete taxonomy of drawback types

A "bullet-proof" methodology has to name every type a broker's clients could qualify for, not just the two or three that are common. Below is the full § 1313 menu. The three bolded rows are the high-volume commercial types that drive ~95% of recovered dollars and are the focus of the Commers matcher.

Cross-cutting statutory mechanics every broker must also apply:

• § 1313(k) — allocation of drawback liability among importer, manufacturer, and claimants.
• § 1313(s) — drawback successor: a successor entity may designate a predecessor's imported merchandise (relevant in M&A; lets an acquirer claim on the acquired book).
• § 1313(u) — the designated import must have been regularly entered or withdrawn for consumption (no TIB, no in-bond-only merchandise).
• § 1313(n) — USMCA / FTA limitation ("lesser of the two duties") — see §4.4.
• § 1313(x) — for destruction, the value of recovered materials (scrap) is deducted from the duty base.

Commers today: the runtime matcher covers the four bolded commercial types — 1313(j)(1), 1313(j)(2), 1313(a), 1313(b). Rejected-merchandise (c) is supported as a broker-curated claim type but is not auto-matched (it turns on defect/non-conformity facts the platform doesn't infer). The specialized types (d, p, g, h, q, e/f) are out of scope for the automated matcher and are handled as manual claims — we surface the eligible duty but the broker builds the claim. This is stated plainly so we never imply "every kind of drawback" is automated; the honest line is "we automate the four types that recover the money, and support the rest as broker-curated claims."

---

3. Eligibility gates (run before any calculation)

A claim is disqualified before math if any gate fails. The matcher applies them in this order:

1. Duty actually paid. The designated import must have paid duty/tax/fee. Duty-free entries have nothing to recover. (Non-zero duty or fees paid.)
2. Consumption entry. The import was entered or withdrawn for consumption (§ 1313(u)); TIB and merchandise never entered for consumption are excluded.
3. Within 5 years. Export/destruction and the claim both fall within 5 years of the import date (§ 1313(r)).
4. Unused, if a (j) claim. For unused-merchandise drawback the goods must not have been used in the U.S. Incidental operations (testing, cleaning, repackaging, inspection, minor manipulation) do not count as "use" (§ 1313(j)(3)); putting the goods into their intended commercial service does.
5. Provision-appropriate. The facts match one type in §2 — you cannot claim unused drawback on goods that were manufactured, or manufacturing drawback without an approved ruling (§7.2).
6. Not previously designated. The import unit has not already anchored a claim (§ 1313(v)).

Commers: gates 1, 3, 6 are automated. Gate 2 is read from entry type where we have it and otherwise flagged for broker confirmation. Gates 4 and 5 are broker attestations — the platform records the assertion and timestamps who made it (audit trail), but "unused" is a factual determination the licensed broker owns.

---

4. Which duties are recoverable (the part that is wrong in most tools today)

This is the section a sophisticated broker will pressure-test hardest in 2026, because the tariff stack has changed faster than most software. Getting it right is the difference between a defensible claim and a penalty.

4.1 Recoverable

• Ordinary Customs duties (HTSUS Column 1 ad valorem / specific rates). ✅
• Merchandise Processing Fee (MPF) and Harbor Maintenance Fee (HMF). ✅ (These are the "fees" in "duties, taxes, and fees.")
• Section 301 (China) duties. ✅ Drawback-eligible.
• Internal-revenue taxes where the specific provision allows (e.g., § 1313(d)).

4.2 NOT recoverable

These are hard legal exclusions, not conservatism — including either in a drawback base produces a false claim.

• Section 232 (steel/aluminum/copper, and derivative) duties. ❌ Barred by the express terms of the Section 232 proclamations themselves — Proclamations 10895 & 10896 (Feb 10, 2025) and 10947 (Jun 3, 2025) each provide that "no drawback shall be available with respect to the duties imposed pursuant to this proclamation." The bar lives in the tariff instrument, not the drawback statute. Where goods carry both 301 and 232, only the 301 portion is recoverable; the 232 portion is excluded. (232 comes back only via a retroactive product exclusion or assessment-error refund — a protest/PSC pathway, not drawback.)
• Antidumping / countervailing duties (AD/CVD). ❌ Categorically excluded by 19 C.F.R. § 190.3(b): "Drawback is not allowable on antidumping and countervailing duties…" This applies to direct-identification claims too, not only substitution. Reinforced by the retrospective nature of AD/CVD — the entry cash deposit is provisional and the final rate is set at liquidation, so there is no fixed "duty paid" to draw back.

Contrast — Section 301 is recoverable (§4.1) precisely because no comparable prohibition exists: neither the § 301 actions nor § 190.3(b) exclude it. The rule is not "exclude all trade-remedy tariffs"; it is "exclude the specific duties an instrument made non-recoverable (232 by proclamation, AD/CVD by regulation), and keep the rest."

4.3 IEEPA "reciprocal" tariffs — current status (as of 2026-07-17)

• In April 2025, CBP confirmed IEEPA reciprocal duties were drawback-eligible.
• On February 20, 2026, the U.S. Supreme Court held that IEEPA does not authorize the President to impose tariffs, invalidating the IEEPA tariff measures. Those duties are being refunded directly through CBP's CAPE (Consolidated Administration and Processing of Entries) pathway in ACE, launched April 2026 — not through the drawback program.
• Methodology consequence: IEEPA amounts must not be included in a drawback duty base for current claims — they are recovered (if at all) via CAPE, and double-recovering the same dollar through drawback would be a false claim. Section 301 and 232 were unaffected by the ruling and follow §4.1/§4.2.

Commers: duty is decomposed by program family at compute time — the recoverable base is base HTSUS + Section 301 only; Section 232/201, IEEPA, and AD/CVD are excluded, read from the duty engine's per-component program family. Both matchers use this eligible base as the 99% basis. For legacy rows imported before the decomposition existed, the matcher falls back to the full duty and stamps the claim as an unverified base — it never silently assumes the whole duty is recoverable. The gross duty stays on the entry so the importer sees gross vs. eligible side by side.

4.4 USMCA / FTA limitation — "lesser of the two duties" (§ 1313(n))

For merchandise exported to Canada or Mexico and entered there for consumption, NAFTA/USMCA caps drawback at the lesser of (a) the U.S. duty paid, or (b) the duty paid to the destination country. Same-condition (unused) substitution drawback is generally not available for USMCA-country exports. This is a hard limit — brokers get audited on it.

Commers (enforced — system control): both matchers read the export's destination country and route USMCA (Canada/Mexico) exports to broker review instead of auto-calculating the full 99%. In the unused matcher a USMCA match is recorded with the pairing and the import's eligible duty as context, but the recoverable amount is left at zero (so it is neither counted nor auto-filed — entry generation requires a positive recoverable) and flagged for USMCA review. The manufacturing matcher excludes USMCA finished-good exports from the auto-claimed quantity and reports the deferred count. A matched export whose destination is blank/unknown is treated the same way (fail-safe — we can't confirm it isn't USMCA) and flagged with a distinct unknown-destination note. We cannot value the §1313(n) cap without the destination-country duty, so the broker applies it — but the platform now guarantees a USMCA (or unconfirmable) export is never silently over-claimed at 99%.

---

5. Matching methodology — import ↔ export

Every claim links a designated import (the duty-paid line) to a qualifying export or destruction. Two bases:

5.1 Direct identification (§ 1313(j)(1), § 1313(a))

The same merchandise is traced from import to export/destruction. Identity is established by lot numbers, serial numbers, or an approved accounting method — FIFO, LIFO, low-to-high, or another GAAP inventory method — applied consistently (19 C.F.R. § 190.14). No HTS substitution is involved; recovery is the actual duty paid on the identified units.

5.2 Substitution (§ 1313(j)(2), § 1313(b)) — the 8-digit HTS standard

TFTEA replaced the old, litigation-prone "commercial interchangeability" test with a bright-line rule: the imported and the substituted merchandise qualify if they share the same 8-digit HTS subheading.

The "other" (basket) exception. If that 8-digit subheading is described as "Other" — i.e., it is a residual basket — the 8-digit match is not enough: the goods must then match at the 10-digit statistical level, and neither may be classified under an "Other" description. This closes the loophole of laundering dissimilar goods through a catch-all subheading.

Commers: the matcher implements exactly this ladder — direct-ID 10-digit (best basis) → substitution 8-digit → substitution 10-digit "other" → unmatched. The "Other" determination is authoritative, not a digit heuristic: it reads the subheading's actual HTSUS description (is it literally "Other"?) rather than guessing from the code; the old "ends in 00" rule survives only as a standalone fallback. The matcher is greedy, best-basis-first: direct-ID beats substitution, and each export and each import unit is consumed at most once (Principle 4). Substitution matching reuses the HTS Classifier (Agent #1) so the 8/10-digit codes are the same ones used on the underlying entries.

---

6. Calculation methodology

6.1 Per-unit averaging (the base for substitution claims)

Duties, taxes, and fees on an entry-summary line are apportioned equally across every unit on that line: per-unit duty = line duty ÷ line quantity. The designated import's recoverable value is the per-unit average of the entry line, not a hand-picked high-duty unit (19 C.F.R. § 190.32(b)). This is what stops a claimant from importing high- and low-value goods on one line and cherry-picking the expensive units for the refund.

Commers: the unused-drawback calculation applies per-unit averaging and a quantity cap — recovery scales to the lesser of exported and imported units at the line's eligible per-unit duty, so exporting 1 unit of a 1,000-unit import recovers one unit's duty, not the whole entry. The eligible duty is first apportioned to the matched line by value share. (Where quantities are absent it falls back to the whole line, capped at the import side.) Unit-tested against the worked examples in §6.5.

6.2 The "lesser-of" rule (substitution only)

For substitution claims, drawback may not exceed 99% of the lesser of:

• (a) the duties/taxes/fees paid on the designated imported merchandise (per-unit averaged), or
• (b) the duties/taxes/fees that would apply to the substituted exported/destroyed article as if it had been imported.

recoverable = 0.99 × min(dutyOnImport, dutyOnExportedArticleAsIfImported)

For same-8-digit substitution at a stable rate the two prongs usually coincide; they diverge when trade-program stacking or rate changes differ between the designated import and the substituted good — which is precisely when the rule bites. Direct-identification claims are not subject to lesser-of; they recover the actual duty on the identified units.

Commers: the calculation evaluates both prongs independently for substitution and takes 99% of the lesser of the two; direct-ID claims skip prong (b) and recover the actual eligible import duty. Prong (b) — the exported article's duty "as if imported" — is computed as its base HTSUS duty (rate resolved from the export HTS via the shared duty engine, applied to the export per-unit value). We use base duty only because the substituted good's country of origin — which would determine its Section 301/232 exposure — is not on the export record; base duty is origin-independent and is the conservative, defensible floor. A broker who supplies the export good's origin can lift prong (b). This reproduces worked example A (§6.5) exactly ($1,188, not $11,088).

6.3 Manufacturing calculation methods (§ 1313(a)/(b); 19 C.F.R. § 190.23)

Manufacturing drawback additionally requires attributing imported input to the exported finished good. The regulation defines four methods; the correct one depends on waste and whether the process yields multiple products:

Destruction claims additionally net out recovered scrap value under § 1313(x).

Commers today: the manufacturing matcher explodes a single-level bill of materials and attributes imported components to exported finished goods with value-apportioned per-line duty, never over-claiming the imported quantity (a conservative used-in / appearing-in hybrid). Not yet automated: multi-level (nested) BOMs, yield/scrap factors (used-in-less-valuable-waste), and relative-value distribution across co-products. Those are broker-validated today and on the roadmap. We disclose this rather than imply full multi-level costing.

6.4 Rejected-merchandise calculation (§ 1313(c))

Recovery is 99% of the duties paid on the specific goods that were defective, non-conforming, or shipped without consent, upon their export or destruction within the 5-year window. Direct identification to the import; no substitution.

6.5 Worked examples

A. Unused substitution, lesser-of binds (§ 1313(j)(2))

• Import: 1,000 units, HTS 8481.80.90xx, entered value $100/unit, duty 3% + Section 301 25% = 28% → $28,000 duty, i.e. $28.00/unit (per-unit average).
• Export: 400 units of same 8-digit HTS, but sourced from a country without Section 301, so "as if imported" duty = 3% only = $3.00/unit.
• Lesser-of: min($28.00, $3.00) = $3.00.
• Recoverable on the 400 exported units: 0.99 × 400 × $3.00 = $1,188.00.
• (Naïve tools that skip prong (b) would wrongly claim 0.99 × 400 × $28 = $11,088 — a ~$9,900 over-claim and a penalty exposure.)

B. Unused direct-ID, no lesser-of (§ 1313(j)(1))

• Same import line, but the exact 400 imported units (FIFO-identified) are re-exported unused.
• Recoverable: 0.99 × 400 × $28.00 = $11,088.00 — the full duty, because direct-ID recovers actual duty paid and lesser-of does not apply.

C. Manufacturing, single-level BOM (§ 1313(a))

• Finished good = 1 imported motor ($40 duty) + 1 imported housing ($10 duty); 2,000 finished units exported.
• Per-unit recoverable input duty = $50; recovery = 0.99 × 2,000 × $50 = $99,000.
• If the process also yielded a valuable scrap co-product, the relative-value method would reallocate part of that $50 base to the co-product → broker review.

---

7. Records & proof (no records, no refund)

7.1 Proof of exportation / destruction

The claimant must establish the date and fact of export or destruction (19 C.F.R. § 190.72–.76). Acceptable proof: bill of lading, air waybill, ocean manifest, AES / Electronic Export Information with ITN, and business records kept in the normal course. Destruction requires evidence of destruction under CBP supervision (or approved after-the-fact). A chronological summary of exports ties each claimed line back to its proof.

7.2 Manufacturing rulings

Manufacturing drawback (a)/(b) additionally requires an approved drawback ruling on file before or with the claim:

• General manufacturing ruling — the claimant adopts a published ruling from Appendix A/B to Part 190 (fast, standardized processes).
• Specific manufacturing ruling — a custom ruling from CBP for a process not covered by a general ruling.

No approved ruling → no manufacturing claim. (Unused-merchandise drawback needs no ruling.)

Commers (enforced — system control): entry generation will not bundle a 1313(a)/(b) entry unless an approved manufacturing ruling is on file (the ruling must be flagged present and carry a non-empty ruling number). A bulk run silently skips manufacturing provisions lacking a ruling (the claims stay eligible/unfiled); an explicit request for a manufacturing provision without a ruling is rejected. Unused-merchandise (j) entries are unaffected.

7.3 Retention

All supporting records: 3 years after liquidation of the claim (19 U.S.C. § 1508; § 190.15). Because liquidation itself can run 1–4 years, practical retention is often 5–7 years end-to-end. Records include the import entry package, proof of duty paid, BOM (manufacturing), inventory/accounting method documentation, and proof of export.

Commers: every claim carries its lineage — the source entry, the duty decomposition, the matched export record, and the match basis — as immutable, timestamped rows. The platform is the system of record that makes the 3-year retention automatic and produces the chronological summary on demand. Proof documents themselves (BOLs, AES) are ingested where available and otherwise referenced.

---

8. Filing methodology (ACE type-47 entry + privileges)

8.1 The claim

Claims are filed electronically in ACE via ABI as a type-47 drawback entry (the electronic successor to paper CBP Form 7551). A complete claim bundles, per § 1313 provision: the designated import entries, the export/ destruction proof, the calculated 99% recoverable, and the claimant/bond data. Required supporting data must reach CBP within 24 hours of filing unless the claimant holds waiver privileges (§8.2).

8.2 Privileges that change the cash-flow and workflow

Accelerated payment is what makes drawback a cash product rather than a multi-year receivable, so a broker's clients will almost always want AP + a sufficient bond.

Commers today: entry generation bundles filable claims by provision into a field-accurate ACE Drawback Entry Summary (the CBP CATAIR/TFTEA Entry Summary record format — see §8.3) and flips the bundled claims to filed. Honest boundary — this GENERATES the transmittable entry; it does not transmit to CBP. Actual ABI transmission requires the broker's certified ABI connection and a valid bond; until that link exists the CBP entry number and claim number remain null. AP/WPN/OTW privileges are broker-held — the platform records which privileges apply to a claim and structures the entry accordingly, but the privileges are applications the broker files with CBP.

8.3 The output file — a field-accurate CATAIR Drawback Entry Summary

The generator does not emit an illustrative or "CATAIR-style" placeholder. It produces a field-accurate ACE Drawback Entry Summary that conforms to the CBP ACE CATAIR — Drawback / TFTEA (Entry Summary) specification, V27 (June 2025) (see §11), the exact fixed-width format ABI accepts:

• 80-character fixed-width records, each with spec-correct field positions and data classes.
• The full transaction record set, in CBP's mandated order: 10 (entry header) → 31 (bond) → 40 / 41 / 42 / 43 (each designated import — line, classification, quantities, revenue) → 50–53 (manufactured-article groupings, for § 1313(a)/(b)) → 60–64 (notices of intent, exam/witness, USMCA) → 70–73 (export/destruction articles and the imports or articles they draw against) → 89 (revenue totals by class) → 90 (grand totals).
• A spec-correct CBP entry number — 3-character filer code + 7-digit sequence + the Appendix-B check digit — with the repeat limits from the Input Structure Map enforced (e.g. ≤ 8 classifications and ≤ 3 quantities per import).
• Wrapped in a B…Y block-control envelope. Deliberately no A/Z batch header/trailer — the batch envelope and ABI password belong to the broker's own certified ABI software.

The result is a file the broker's licensed ABI software ingests and transmits to CBP directly — not a report to be re-keyed. That is what turns "the broker files, Commers produces the claim" into a real handoff: Commers builds the exact records CBP expects, and the broker's ABI session and bond carry them the last mile.

8.4 The honest filing story for a broker prospect

"Commers is the matching, calculation, and entry-generation engine. It produces a CBP-ready type-47 drawback entry with a fully reconciled 99% figure and the complete record trail behind it. The licensed broker transmits it through their own certified ABI connection and their bond — which means the broker stays the filer of record and the party CBP deals with. That is a feature, not a gap: it keeps the compliance liability and the customer relationship with the broker, and it means Commers never has to be an ABI-certified filer to deliver the recovery."

This is the direct answer to the two questions brokers press hardest on — ABI filing and audit defense: the broker files and defends; Commers produces the defensible claim and holds the evidence.

---

9. Audit-defense methodology

CBP can question a filed claim for years after payment (accelerated payment is conditional on later liquidation). A claim is audit-defensible when, for every recovered dollar, the platform can reproduce on demand:

1. The designated import — entry number, line, HTS, entered value, duty decomposition (and the exclusion of 232/AD-CVD/IEEPA from the base).
2. The match basis — direct-ID accounting method (FIFO/etc.) or the exact 8-/10-digit substitution ladder that qualified it.
3. The calculation — per-unit averaging, lesser-of prongs (both values shown), and the manufacturing method + BOM where applicable.
4. The proof of export/destruction — BOL/AES/ITN, tied by chronological summary.
5. The provenance — who (broker vs. AI) asserted each judgment call, and when — the immutable audit trail.
6. The non-duplication proof — that the designated units were never claimed twice (§ 1313(v)).

Commers: items 1–3, 5, 6 are produced automatically from the claim's stored lineage; item 4 is assembled from ingested/referenced export proof. The design principle from the platform — broker outranks AI, everything is logged — is the audit-defense story: a claim is never a black-box number, it is a reproducible chain from CBP entry to refund.

---

10. Assumptions, conservatism, and known limitations

Every conservative choice in one place, each pointing back to where it is argued in full rather than re-argued here:

• Conservative duty base — excludes Section 232, AD/CVD, and IEEPA and caps at 99% of the lesser prong for substitution; the platform under-claims before it over-claims (§4.2, §4.3, §6.2).
• Substitution standard — the TFTEA 8-digit HTS standard plus the "other" 10-digit exception; no legacy commercial-interchangeability analysis (§5.2).
• Manufacturing scope — single-level BOM today; multi-level BOMs, yield/scrap factors, and relative-value co-product distribution are broker-validated and on the roadmap (§6.3).
• Human in the loop — "unused" determinations, USMCA-destination caps, rejected-merchandise facts, and manufacturing-ruling coverage are broker attestations, recorded and timestamped (§3, §4.4, §7.2).
• No transmission (yet) — the platform generates the type-47 entry; the broker transmits via certified ABI, and CBP entry/claim numbers stay null until then (§8).
• Law is live — tariff eligibility tracks current law; the IEEPA line reflects the Feb 20, 2026 Supreme Court ruling and must be re-reviewed if the landscape shifts again (§4.3).

---

11. Source register

Primary authority (control in this order):

• 19 U.S.C. § 1313 — Drawback and refunds
• 19 C.F.R. Part 190 — Modernized Drawback (TFTEA)
  • Subpart B — Manufacturing Drawback (§ 190.22 substitution, § 190.23 calculation methods)
  • Subpart C — Unused Merchandise Drawback (§ 190.32 substitution calc + lesser-of)
  • Subpart I — Waiver of Prior Notice; Accelerated Payment
• 19 C.F.R. Part 191 — legacy Drawback (pre-TFTEA claims only)
• 19 U.S.C. § 1508 — Recordkeeping

CBP filing specification (the ACE type-47 record format the generator emits, §8):

• CBP — ACE CATAIR: Drawback Claim (Entry Summary) Transaction, TFTEA — V27 (June 2025) — the 80-character fixed-width ABI record specification the generated entry file conforms to (§8.3).
• CBP Form 7551 — Drawback Entry (the paper predecessor to the electronic type-47 entry).

CBP guidance:

• CBP — Drawback in ACE and Drawback ACE FAQs
• CBP — What Every Member of the Trade Community Should Know About: Drawback (ICP)
• CBP — IEEPA FAQ and Section 232 FAQ

Legislation:

• Trade Facilitation and Trade Enforcement Act of 2015 (TFTEA), Pub. L. 114-125 — the modernization that set the uniform 5-year window and the 8-digit substitution standard.

---

Maintained alongside the Drawback Advisor agent. When the calculator changes, this document and Drawback Advisor change with it — the methodology is the contract the code implements.