HTS Code Lookup API vs Manual Classification: Cost and Accuracy
What changed: GingerControl's 99.86% internal-test figure is now stated at the 10-digit level.
An HTS code lookup API and manual classification answer different questions. A lookup API, such as the USITC's free hts.usitc.gov/reststop API, returns schedule lines and rates for a keyword or code range; a classification API returns a code for a product description; manual classification is a person applying the GRI and the notes, at 30 minutes to 2 hours per product by GingerControl's estimate. For an importer adding 2,000 SKUs a year, the lookup API is free plumbing, a classification API saves research hours, and a person still reviews the codes that matter.
In this post
- How does an HTS code lookup API compare to manual classification?
- How does the USITC HTS REST API work?
- What does manual HTS classification cost?
- Which is more accurate, an API or a person?
- When does manual classification still make sense?
- What does a classification API add over a lookup API?
- Wiring HTS data into your catalog
- References
An HTS code lookup API lets software query the Harmonized Tariff Schedule instead of a person searching it by hand; a classification API goes further and suggests the code for a product. GingerControl's Classification + Tariff Calculation API is the second kind: one call returns a 10-digit HTS code and its U.S. duty layers, for enterprise customers after a call. I'm Chen Cui, a GingerControl co-founder; this post explains both kinds of API, including the free government one, and when manual work is still the right call.
How does an HTS code lookup API compare to manual classification?
A lookup API is faster than a person at finding lines in the schedule but does not decide anything; a classification API suggests a code but needs review; manual classification decides slowly and leaves whatever record the classifier writes.
| Lookup API (USITC reststop) | Classification API | Manual classification | |
|---|---|---|---|
| Input | A keyword or a code range | A product description, sometimes origin | The product, its spec sheet, samples |
| Output | Schedule lines, rates, footnotes | A suggested code; some APIs add duty or reasoning | A code and, if written down, the reasoning |
| Decides the heading? | No | Suggests it | Yes |
| Cost | Free, no key | Per call or by contract | Staff time: 30 minutes to 2 hours per product, by our estimate |
| Record for audit | None beyond the query | What the response contains | Whatever the classifier writes |
Sources: [1]. For a side-by-side of the free tools and paid lookup tools, see HTS code lookup tools compared; for step-by-step use of each free government source, see the best free HTS code lookup tools.
How does the USITC HTS REST API work?
The USITC's HTS REST API is a public JSON service at hts.usitc.gov/reststop that searches and exports the schedule, with no account or key [1]. Three endpoints cover most needs:
GET https://hts.usitc.gov/reststop/search?keyword=vacuum%20flask
GET https://hts.usitc.gov/reststop/exportList?from=9617&to=9618&format=JSON&styles=false
GET https://hts.usitc.gov/reststop/releaseList
- search returns rows with
htsno,description,indent,units,general,special,other(column 2),footnotes,quotaQuantityandadditionalDuties. - exportList returns a code range as JSON, CSV or XLSX; on October 11, 2026 it returned an error unless
styles=trueorstyles=falsewas included, although the guide listsstylesas optional [3]. - releaseList returns the revision history; on October 11, 2026 it showed Revision 21 of the 2026 HTS, effective October 9, 2026, after 32 revisions in 2025 [1].
The USITC's user guide describes the same service as a RESTful web service API with the base URL hts.usitc.gov/reststop [3]. The guide says search returns "up to the first 100" results [3], but a search for "steel" on October 11, 2026 returned 403 rows, so do not rely on that cap. The API has no country-of-origin input and does not total duties; Section 301 and 232 duties sit in Chapter 99 headings, not on your product's line.
What does manual HTS classification cost?
Manual classification costs research time: by GingerControl's estimate, 30 minutes to 2 hours per product, and 2 hours or more for a compliance-grade report. At the May 2025 BLS mean wage for compliance officers, $42.50 an hour [2], that is about $21 to $85 in wages per product before benefits and overhead, by our arithmetic.
| New products a year | Manual research at 1 hour each | Wages at $42.50 an hour |
|---|---|---|
| 500 | 500 hours | $21,250 |
| 2,000 | 2,000 hours | $85,000 |
| 10,000 | 10,000 hours | $425,000 |
Worked example; the 1 hour per product is our assumption, within GingerControl's estimate of 30 minutes to 2 hours. Add rechecks: the schedule was revised 32 times in 2025 [1]. The full ROI model, with review time and tool cost, is in the ROI of automated HTS classification, and why the hours pile up as a catalog grows is in why manual HTS classification doesn't scale.
The manual classification research at catalog scale analysis shows why adding SKUs eventually requires a different research workflow.
Quotable insight: Teams compare "API vs manual" as if the API replaced the person. It replaces the search, not the judgement. A lookup API saves the minutes spent finding lines; a classification API saves the hour spent deciding between them; neither saves the few minutes a responsible person should spend checking the code before the code goes on an entry, and that review is the cheapest part of the process.
Which is more accurate, an API or a person?
Neither has a published head-to-head benchmark, so test both on your own products. Vendor figures measure different things (a private test set, a reclassification rate, a target), as compared in HTS classification providers compared on accuracy. Score at 6, 8 and 10 digits separately: CBP's tariff classification guide says the legal text extends to 8 digits and the last two are statistical [4]. GingerControl's quick classification line, which its API uses, scored 99.86% at the 10-digit level on GingerControl's own internal test of 1,000+ products; that is a first-party test, not an independent benchmark.
Review what GingerControl's 99.86% classification accuracy measures before comparing that first-party result with another vendor's benchmark.
When does manual classification still make sense?
Manual classification still makes sense for products new to you, goods whose heading turns on a lab analysis, and codes with large duty at stake and real doubt, where a binding ruling is worth asking for. Rulings are free [7], and CBP says its National Commodity Specialist Division generally issues them within 30 calendar days [5]. CBP's reasonable care checklist also asks whether "a laboratory analysis or other specialized procedure is suggested" for your goods [6].
What does a classification API add over a lookup API?
A classification API adds the decision step, and some also return the duty. GingerControl's Classification + Tariff Calculation API takes a product description and origin and returns a 10-digit HTS code with its U.S. duty layers in force when called (General, Special, Section 301, Section 232 metals, other Chapter 99 lines), 120 items per call by default, with split-code composite duty such as Chapter 91 watches returned as components. It is for enterprise customers after a call; there is no public price. GingerControl's classifications are for general reference, educational, and planning purposes; before a code goes on an entry, your licensed customs broker reviews it. GingerControl is not a customs broker, does not file entries, and does not give legal advice.
Other vendors' APIs are compared in the best HTS classification APIs; building your own instead is weighed in build vs buy for HTS classification; and the legal limits on giving codes to others, including CBP's view that a tool is not a person [8], are in can AI tools legally classify HTS codes?.
Wiring HTS data into your catalog
Use the free USITC API for schedule text and rates; when your catalog needs a code and its U.S. duty for every new SKU, the Classification + Tariff Calculation API returns both in one call. See what the API returns for your SKUs
An HTS classification API fits between the product record and the approved classification stored by the receiving system.
References
- U.S. International Trade Commission, HTS REST API at hts.usitc.gov/reststop (read October 11, 2026)
Data cited: search, exportList and releaseList endpoints with no login or key; exportList returned HTTP 400 without the styles parameter; search fields htsno, description, indent, units, general, special, other, footnotes, quotaQuantity, additionalDuties; export as JSON, CSV or XLSX; a search for "steel" returned 403 rows; Revision 21 of the 2026 HTS effective October 9, 2026; 32 revisions in 2025
Source: USITC HTS release list - U.S. Bureau of Labor Statistics, Occupational Employment and Wages, May 2025, SOC 13-1041 Compliance Officers
Data cited: mean hourly wage $42.50
Source: BLS OEWS, Compliance Officers - U.S. International Trade Commission, HTS external user guide (last updated June 15, 2023; read October 11, 2026 in the Internet Archive copy of September 17, 2026, because usitc.gov blocked direct reading)
Data cited: the HTS system "provides an interface to extract search and export results" and "uses a Representational State Transfer (RESTful) web service API to provide this data"; base URL hts.usitc.gov/reststop; search returns "up to the first 100 tariff articles"; exportList takes format, from, to and styles (listed as not required)
Source: USITC HTS external user guide - U.S. Customs and Border Protection, "Tariff Classification" informed compliance publication (May 2004)
Data cited: "The legal text of the HTSUS extends only up to the 8-digit level"; 10-digit numbers are statistical
Source: CBP "Tariff Classification" ICP - U.S. Customs and Border Protection, eRuling requirements page
Data cited: "Generally, the NCSD will issue all such rulings within 30 calendar days of the date of receipt"; up to five items of the same class or kind
Source: CBP eRuling requirements - U.S. Customs and Border Protection, Reasonable Care informed compliance publication (September 2017 revision)
Data cited: classification question on laboratory analysis or other specialized procedures
Source: CBP Reasonable Care publication - U.S. Customs and Border Protection, Rulings Program informed compliance publication (September 20, 2023)
Data cited: "There are no charges or fees associated with the rulings program"
Source: CBP Rulings Program ICP - U.S. Customs and Border Protection, ruling HQ H350722 (January 16, 2026)
Data cited: "A tool does not constitute a 'person' as defined by 19 C.F.R. § 111.1"; the classification decision "must be made by a duly licensed customs broker"
Source: CBP CROSS, HQ H350722
Questions people ask
How do I use the USITC HTS reststop API to search by keyword?
Send a GET request to hts.usitc.gov/reststop/search with a keyword parameter, for example keyword=vacuum flask. On October 11, 2026 it answered without a login or key and returned JSON rows with the HTS number, description, indent, units, general, special and column 2 rates, footnotes and additional duties. For a developer building a 2,000-SKU lookup, cache results and record the revision from /reststop/releaseList.
Is there documentation for the USITC HTS REST API?
Yes. The USITC's external user guide for the HTS describes a RESTful web service API with the base URL hts.usitc.gov/reststop, for extracting search and export results. For a developer integrating it into a product catalog of 5,000 items, test the search, exportList and releaseList endpoints directly; one claim in the guide, a 100-result cap, did not hold when we searched in October 2026.
Can the USITC API classify a product description?
No. It searches the tariff schedule's own text and exports code ranges; it does not decide which heading fits your product or total the duty for a country of origin. For an importer adding 300 products a year, a classification API or a person still has to apply the GRI and the Section and Chapter Notes to reach the code.
How much does manual HTS classification cost compared with an API?
Manual work costs time: by GingerControl's estimate, 30 minutes to 2 hours per product, which at the May 2025 BLS mean wage for compliance officers ($42.50 an hour) is about $21 to $85 in wages per product, before overhead. For a platform classifying 2,000 SKUs a year, compare that with the API's price plus the time a person spends reviewing each code it returns.
Is an HTS classification API more accurate than manual classification?
No shared benchmark answers that, and vendor figures measure different things. For an importer with 1,000 SKUs, test both on 50 items your broker already confirmed, scoring 6, 8 and 10 digits separately. GingerControl's quick classification line, which its Classification + Tariff Calculation API uses, scored 99.86% at the 10-digit level on GingerControl's own internal test of 1,000+ products.
How long does manual HTS classification take per product?
By GingerControl's estimate, 30 minutes to 2 hours per product, and 2 hours or more for a report fit for a compliance file. Simple, single-material goods can take minutes. For a team adding 40 composite SKUs a month, that is 20 to 80 hours of research before review; an API or research tool cuts the research, not the review.
Can an API's HTS code go straight onto a customs entry?
Not safely. The importer of record files the classification using reasonable care, and when the code is for someone else's goods, CBP said in HQ H350722 that a tool is not a person and a licensed broker must make the decision. For a platform serving 200 importers, keep a broker review step between the API response and the entry.
When should we still classify manually?
Classify manually, or ask CBP for a binding ruling, when the product is new to you, when a lab analysis decides the heading, or when the duty at stake is large and the code is uncertain. For an importer launching five products of the same class or kind, one free eRulings request can cover all five, and CBP says its specialists generally answer within 30 calendar days.
Classifying SKUs one ruling at a time?
Hand this to GingerOS.
Pick a 30-minute slot on the next page with Chen Cui, our CEO. Bring one notice or one SKU list (an ERP export, a spreadsheet or an email will do) and see what GingerOS does with it.
Reviewed by Michael Weick, LCB / CCS




