Import Cost Methodology

1. What the AI does — and does not do

AI (a large language model) is used in exactly two places: to understand your product description (extract attributes like material and primary use, and rewrite your words into tariff-schedule vocabulary) and to summarize public CBP ruling text shown alongside candidates. Every AI-generated item is labeled “AI-generated” in the results.

The AI never picks your tariff rate, never modifies the official HTS titles, and never performs the calculation. The final rate always comes from the official tariff schedule and rules stored in our database, and every number is computed deterministically by code.

2. How HTS candidates are found

We search the current US Harmonized Tariff Schedule (about 24,000 active nodes) using three independent retrieval channels, then merge their results:

• Lexical search — keyword scoring over official titles, aliases and generated everyday vocabulary.
• Semantic (embedding) search — vector similarity between your description and each heading's search document.
• Attribute recall — structured attributes (material, primary use, construction, …) mapped to HTS concepts.

Merged candidates are scored with a hierarchy-aware ranking (lexical fit, attribute fit, position in the schedule, specificity, semantic score, and penalties for conflicting attributes). We output the top 3–5 candidates, each with a qualitative confidence level (High / Medium / Low) — never a made-up percentage. When the description is too vague to distinguish between candidates, we ask one clarifying question at a time (at most three).

3. Which tariff rules apply

Each estimate is built line by line from rules stored with a source URL, effective dates and a rule version. For the current rule set:

• Base duty — the HTS General (Column 1) rate of the confirmed subheading.
• Section 301 list duties — China-origin goods, resolved per HTS line against official list membership: the legacy lists (List 1–3 at 25%, List 4A at 7.5%, plus heading 9903.88.04 at 25%) and the 2024 four-year review tranches (25%, 50% and 100% on the lines it named, some keyed to a 10-digit statistical reporting number). A line that is on no list bears no Section 301 duty; there is no blanket fallback. Where the four-year review raised a line’s rate it replaced the legacy rate rather than adding to it. Product exclusions are not evaluated, so a line whose exclusion was granted is overstated here.
• Section 301 forced-labor duty — 12.5% on products of China under the 2026 action, effective 24 July 2026. The action’s Annex I/II product exceptions are not evaluated: the duty is applied at its published default scope.
• Section 122 temporary surcharge — the economy-wide 10% surcharge that applied from 24 February 2026 until it expired on 24 July 2026 (historical dates only).
• IEEPA and reciprocal duties — the China IEEPA tariff (10% → 20% → 10%) and the reciprocal tariff (34% → 84% → 125% → 10% pause), both terminated on 20 February 2026. These appear only for import dates when they were in force.
• Merchandise Processing Fee (MPF) — formal-entry fee with the CBP annual minimum/maximum clamp.

The Harbor Maintenance Fee (HMF) applies to ocean shipments only and is not included in the estimate. Rules are selected by your import date: where a rule changed over time, the version effective on that date is the one applied — a 2025 import date resolves to the duties that were in force in 2025, not today’s. Section 232 duties are not modeled.

One limitation to know about: Section 301 list membership is stored as it stands in the current HTS revision, and the rate windows are historical. A very old import date therefore resolves against today’s list membership with that period’s rate, rather than a reconstruction of the list as it was printed that week. For 2025 and 2026 dates this makes no difference; treat pre-2025 comparisons as indicative only.

4. How landed cost is calculated

base duty       = customs value × HTS general rate
additional duty = customs value × each applicable additional rate
MPF             = customs value × 0.3464%, clamped to the CBP annual min/max
total duty      = base duty + additional duties
landed cost     = product cost + freight + insurance + total duty + fees
                  + other import costs
per unit        = landed cost ÷ quantity

Customs value and product cost are separate inputs: the customs value is what duties are assessed on, the product cost is what you paid your supplier, and both feed different totals. Landed cost uses the product cost; duty uses the customs value. All money math is exact decimal arithmetic (no floating point), computed at 4-decimal precision internally and displayed rounded to 2 decimals. Every component is rounded separately before it is summed, so a line’s amount is reproducible from the rate shown next to it.

Freight, insurance, broker, port and other import costs are counted when you provide them. A field you leave empty counts as zero and is labeled as not provided — it is never estimated for you.

5. Reproducibility and versioning

Every calculation is stored as a snapshot containing the exact inputs, the HTS revision used, the rule versions applied, the engine version and a timestamp. The same inputs, HTS revision and rule versions always produce the same output — results can be re-computed and audited after the fact.

6. Data sources and freshness

• US Harmonized Tariff Schedule (USITC) — ingested September 2026, HTS revision recorded with every calculation. See Data Sources.
• CBP rulings — public ruling texts, retrieved September 2026, shown as reference evidence only.
• Trade-remedy and fee rules — CBP/USTR guidance with per-rule source URLs and effective dates.