Automated Tariff Code Classification assigns HS and HTS tariff codes to products at scale using predefined logic and classification models. It helps businesses standardize product classification across large catalogs without manual intervention.

Core capabilities

Automatically classify large product catalogs using standardized logic.

High-volume classification

Apply consistent tariff codes across systems and transactions.

Consistent code assignment

Reduce reliance on manual classification workflows.

Reduced manual effort

Support growing catalogs and global expansion.

Scalable classification workflows

Integrate classification workflows through APIs by submitting product attributes and retrieving assigned tariff codes. Setup typically includes mapping product catalog data and configuring classification rules within your system.

Getting started

Integrate via REST APIs to submit product descriptions and receive tariff code assignments in real time.

How it integrates

What Automated Tariff Code Classification solves

When to use Automated Tariff Code Classification

Automated Tariff Code Classification

Get Started

Avalara Item Classification automates the complex process of determining HS codes, taking into consideration country, product attributes, and other factors.

Apply Avalara-maintained tax rules and rates tailored to your product category and jurisdictions.

Industry-specific logic

Support tiered pricing, bundled offerings, exemptions, and multi-jurisdiction calculations without custom engines.

Complex scenarios

Stay current with evolving requirements through regularly updated content and configuration options.

Regulatory alignment

Integrate with billing, ERP, and custom apps using REST APIs built for high-volume transaction processing.

Operational scale

Use API documentation, integration checklists, and environment setup guides linked from each product below. Provisioning and access details depend on your Avalara products and onboarding path.

Avalara APIs provide REST interfaces for billing, subscription, ERP, and custom applications. They support scalable tax determination and high-volume processing so you can embed compliance into your existing workflows.

Tariff Code Classification

Self-Serve Tariff Code Classification allows users to manually classify products by entering product details and attributes. It is designed for interactive classification workflows when automation is not required.

Classify products using descriptions and attributes.

Interactive classification

Flexible input handling

User-controlled decisions

Enable interactive classification through APIs that accept product descriptions and return tariff code suggestions. Implementation requires passing product data and handling user-driven selection workflows.

Use the self-serve interface directly or integrate via APIs for a hybrid classification approach.

Self-Serve Tariff Code Classification

Managed and Premium Tariff Code Classification provide expert-led and hybrid classification services to support accurate tariff code assignment for products. These solutions are designed for businesses that require high confidence in classification, especially for complex or compliance-sensitive scenarios.

Leverage specialists to assign and validate tariff codes for accuracy and compliance.

Expert-led classification

Combine automated classification with expert review to support high-confidence results.

Hybrid classification model

Ensure classification outputs meet regulatory requirements and support audit workflows.

Compliance and audit support

Support classification for products with complex attributes, regulatory requirements, or high-risk profiles.

Complex product handling

Get started by onboarding products into classification workflows.

You can integrate APIs to submit product data and retrieve validated tariff codes, with support for automated processing and expert review.

Submit product data through APIs or batch uploads and receive expert-assigned tariff codes with supporting documentation.

Managed and Premium Tariff Code Classification

Avalara Trade Compliance helps businesses manage regulatory requirements for cross-border transactions. It supports compliance with trade rules, restrictions, and regulatory obligations.

Apply trade compliance rules across jurisdictions.

Regulatory rule management

Risk reduction

Help ensure transactions meet regulatory requirements.

Cross-border control

Integrate compliance checks into transaction workflows using APIs to validate cross-border transactions against regulatory requirements. Setup includes configuring jurisdiction rules and adding compliance checks to order processing workflows.

Integrate via REST APIs to screen transactions and retrieve compliance status in real time.

Avalara Trade Compliance

The Trade and Tariff Library provides access to global tariff schedules, duty rates, and trade data required for cross-border compliance and decision-making.

Access global tariff schedules and duty rate information.

Centralized tariff data

Stay aligned with changing trade regulations.

Up-to-date trade data

Support classification and compliance workflows.

Data-driven decisions

Access tariff and trade data through APIs by querying country, product, and classification inputs. Integration involves incorporating tariff data retrieval into classification, calculation, or compliance workflows.

Query tariff data and duty rates via REST APIs to embed trade data into pricing, compliance, and logistics workflows.

Avalara Trade and Tariff Library

SECTION XI: TEXTILES AND TEXTILE ARTICLES	Chapter 61: Articles of apparel and clothing accessories, knitted or crocheted	T-shirts, singlets and other vests, knitted or crocheted.	Of cotton	Of cotton	Men's or boys':	Other T-shirts

Stylish Back Printed 3/4th Sleeve T-Shirt

LEWEL Men's Stylish Back Printed 3/4th Sleeve T-Shirt (Black)

A list of product attributes as key/value pairs.
For best classification quality, send commonly used attribute names exactly as:
`material`, `price`, `weight`, `gender`, `dimensions`, or `length`/`width`/`height`.
Other attribute names can also be sent as additional fields.


The country of destination. This field is required if `depth = full`. This field expects a two-character ISO 3166 country code.

A URL that points to an image of the product.

Primary product text used for classification. This should clearly describe what the product actually is.
Include crisp and specific details such as product type, key material, specifications, and intended use where available.
A clear and detailed title materially improves classification quality.


The product URL that includes full details about the product.

Context

Data quality evaluation results.
Only returned when the `evaluate_dq=true` query parameter is specified.


Quality label derived from dqScore:
- "Good": dqScore ≥ 0.8
- "Fair": 0.6 ≤ dqScore < 0.8
- "Poor": dqScore < 0.6


List of attributes successfully extracted from the product description.

List of attributes deduced from keywords/context rather than explicitly
present in the product description (e.g. "DVD" implies material "optical").
Distinct from foundAttributes.


List of required attributes that were not found in the product description.

Prediction of Hscode along with additional compliance message

A validation message for the hscode provided by customer

The enriched context of a product, with required enhanced_field.

* `SubscriptionError`: The user is not entitled to use the requested prediction mode.


An attribute successfully extracted from the product description.

The attribute name (e.g., "material", "gender", "color").

HTTPValidationError

An attribute deduced from keywords/context rather than explicitly present in the product description.

The inferred attribute name (e.g., "material", "form").

* `InternalError`: There is an internal error in the server.


* `InvalidDepth`: The depth parameter is not supported.
* `MissingCOD`: The country of destination (COD) is missing for a full-code prediction.


A detailed message about the JSON error response.

A required attribute that was not found in the product description.

A prediction of a Harmonized System code or an Avalara tax code.

SECTION XI: Textiles And Textile Articles	Chapter 61: Articles Of Apparel And Clothing Accessories, Knitted Or Crocheted	T-shirts, singlets and other vests, knitted or crocheted.	Of cotton	Of cotton	Men's or boys':	Other T-shirts

The predicted Harmonized System code or Avalara tax code.

The level of confidence. Accepted values are HIGH, MID, and LOW.

HIGH - Codes that have historically had an accuracy rate of 85% or higher are classified as "HIGH".

MID  - Codes that have historically had an accuracy rate between 75% and 85% are classified as "MID".

LOW  - Codes that have historically had an accuracy rate below 75% are classified as "LOW".


Data quality evaluation results (only present when evaluate_dq=true query parameter is used).

A description of the predicted Harmonized System code or Avalara tax code.

* `InvalidAuthHeader`: The provided authorization header is invalid.


The location where the validation error occured.

A detailed message about the validation error response.

ValidationError

# Access to the Production API
API access requires a bearer token, which can be obtained using your Avatax account ID and license key.

To acquire a bearer token:
1. [Create a License Key if not already created](https://knowledge.avalara.com/bundle/nps1728373667883_nps1728373667883/page/kys1728459996876.html).
2. Call Avalara Identity to get an access token (choose based on your endpoint URL):

<details open>
<summary><strong> hscode.api.avalara.com</strong> — scope: <code>hs-code-classification</code></summary>

```bash
curl -X POST https://identity.avalara.com/connect/token \
-H "Content-Type: application/x-www-form-urlencoded" \
-d "grant_type=client_credentials&client_id=<avatax_account_id>&client_secret=<license key>&scope=hs-code-classification"
```

Sample response:

```json
{
   "access_token": "<jwt_access_token>",
   "expires_in": 3600,
   "token_type": "Bearer",
   "scope": "hs-code-classification"
}
```

</details>

3. The `access_token` can be used to interact with the API.
4. The `access_token` is valid for 1 hour, after which you will need to request a fresh token.

# Access to the Sandbox API
API access requires a bearer token, which can be obtained using your Avatax account ID and license key.

To acquire a bearer token:
1. [Create a License Key if not already created](https://knowledge.avalara.com/bundle/nps1728373667883_nps1728373667883/page/kys1728459996876.html).
2. Call Avalara Identity to get an access token (choose based on your endpoint URL):

<details open>
<summary><strong> hscode.api.sbx.avalara.com</strong> — scope: <code>hs-code-classification</code></summary>

```bash
curl -X POST https://ai-sbx.avlr.sh/connect/token \
-H "Content-Type: application/x-www-form-urlencoded" \
-d "grant_type=client_credentials&client_id=<avatax_account_id>&client_secret=<license key>&scope=hs-code-classification"
```

Sample response:

```json
{
   "access_token": "<jwt_access_token>",
   "expires_in": 3600,
   "token_type": "Bearer",
   "scope": "hs-code-classification"
}
```
3. The `access_token` can be used to interact with the API.
4. The `access_token` is valid for 1 hour, after which you will need to request a fresh token.

Note: The sandbox API is only available for a fixed number of classification requests.

**Important**: Unlike Production, Sandbox access is not automatically provisioned via Salesforce.
Having a valid Avatax account ID and license key lets you authenticate and obtain a token,
but your account must be separately provisioned for the Sandbox environment before API calls
will succeed — otherwise you will receive a 403 Forbidden response.

To request Sandbox provisioning, Contact Go Live Team.


API service to classify products into Harmonized System Codes (HS Codes)

HSCode classification service

Predicts the HS code along with additional compliance prediction.

The length of the HS code to be predicted. `full` = 10 digits; `6` = 6 digits.

depth

Depth

api_version

API Version

validate_hscode

validate hscode

Enable data quality evaluation. When true, response includes `dataQuality`
Returns quality score, label, found/missing attributes


evaluate_dq

Data Quality Evaluation

EnrichedContext

Predicts the HS code using the provided product information for the specified country of destination.

**Performance and timeout guidance**
- `standard`: ~300 ms (typical).
- `enhanced_quick`: response time can vary from near-standard latency to ~8-12 seconds (p95 observed for slower requests).
- `enhanced_precise`: response time can vary from near-standard latency to ~25-30 seconds (p95 observed for slower requests).
- Actual latency varies by request and selected mode; not every request in an enhanced mode will be at the upper range.

**Client timeout recommendation**
- Configure your client timeout to at least **60 seconds** when using enhanced modes.

**Batch size limit**
- Send a maximum of **10 products per request** (applies to all prediction modes).


Customer-facing prediction strategy:
- `standard`: Default path; optimized for speed and efficiency.
- `enhanced_quick`: Enhanced path designed for improved capability (requires entitlement).
- `enhanced_precise`: Enhanced path designed for higher accuracy and quality (requires entitlement).
If entitlement is missing, enhanced modes return a forbidden response.


prediction_mode

Prediction Mode

Enable data quality evaluation. When true, response includes `dataQuality` field
Supported for all prediction modes (standard, enhanced_quick, enhanced_precise).
Returns quality score, label, found/missing attributes


Contexts

Predicts the HS code using the provided product information.
- `prediction_mode=standard` routes to the smart HS6-aware full-code path.
- `prediction_mode=enhanced_quick` / `enhanced_precise` route to the enhanced ICP path (requires entitlement).
If entitlement is missing, enhanced modes return a forbidden response.

**Performance and timeout guidance**
- `standard`: ~300 ms (typical).
- `enhanced_quick`: response time can vary from near-standard latency to ~8-12 seconds (p95 observed for slower requests).
- `enhanced_precise`: response time can vary from near-standard latency to ~25-30 seconds (p95 observed for slower requests).
- Actual latency varies by request and selected mode; not every request in an enhanced mode will be at the upper range.

**Client timeout recommendation**
- Configure your client timeout to at least **60 seconds** when using enhanced modes.

**Batch size limit**
- Send a maximum of **10 products per request** (applies to all prediction modes).


Customer-facing prediction strategy:
- `standard`: Default smart HS6-aware path.
- `enhanced_quick`: Enhanced path designed for improved capability (requires entitlement).
- `enhanced_precise`: Enhanced path designed for higher accuracy and quality (requires entitlement).
If entitlement is missing, enhanced modes return a forbidden response.


Smart predict HS Code (batch) from provided HS6 code

Predicts a list of possible HS code using the provided product information for the specified country of destination.

Success. Returns an outer array with one entry per input product.
Each entry is an inner array of candidate HS codes ranked by confidence.


Automated Tariff Code Classification

MissingAttribute