{"id":1041,"date":"2023-05-01T10:01:10","date_gmt":"2023-05-01T10:01:10","guid":{"rendered":"https:\/\/artelnics.com\/?post_type=case_studies&p=1041"},"modified":"2026-08-21T20:20:26","modified_gmt":"2026-08-21T20:20:26","slug":"hscodes","status":"publish","type":"case_studies","link":"https:\/\/artelnics.com\/case_studies\/hscodes\/","title":{"rendered":"Automation of HS codes assignment for Nestle"},"content":{"rendered":"
\n
An increasing number of businesses are choosing to automate their manual processes. This allows them to streamline their workflows, become more productive, cut costs, and reduce human error. Manual data entries fall victim to human error. Moreover, they consume a lot of time, which could be used for other, more complex tasks. This is why a Nestle decided to invest in automating the HS assignment process for each of its products.<\/section>\n
\n
\n

HS codes<\/h2>\n

The HS code is an international standardized system of numbers to classify traded products. In an international shipment, every item must have an HS code and should be placed in the Commercial Invoice. The code must be correctly introduced to determine the number of import duties or other taxes the company must pay.<\/p>\n

In the following graph, we can observe how the HS codes are organized:<\/p>\n

\"HS
HS code hierarchy used to classify traded products.<\/figcaption><\/figure>\n

As we can see, each pair of numbers refers to some specific information about the product.<\/p>\n<\/section>\n

\n
\n

Machine learning algorithm<\/h2>\n

This project aims to create a machine learning model to automatically assign the HS code to each product based on its description. By using this algorithm, it is possible to eliminate manual data entry. This simplifies the work of warehouse employees, reduces errors in assigning HS codes to products, and saves time and costs.<\/p>\n

The algorithm consists of a machine learning model that describes a product and assigns the proper HS code by using specific rules. We can observe this process in the following diagram:<\/p>\n

\n
01<\/span>Product descriptions<\/strong><\/p>\n

Millions of labelled descriptions are prepared for training.<\/p>\n<\/div>\n

02<\/span>Machine learning model<\/strong><\/p>\n

The model learns the patterns that connect text with HS codes.<\/p>\n<\/div>\n

03<\/span>Unique rules<\/strong><\/p>\n

Relevant words and combinations are converted into assignment rules.<\/p>\n<\/div>\n

04<\/span>Automatic code<\/strong><\/p>\n

New products receive the most appropriate HS code with fewer manual errors.<\/p>\n<\/div>\n<\/div>\n

First, the algorithm is trained using millions of product descriptions and their matching HS codes, employing machine learning. Then, it is possible to build a mathematical model that consists of specific rules for the HS code assignment. These rules are a single word or a combination of terms contained in the product\u2019s description and must be unique for each product.<\/p>\n<\/section>\n

\n
\n

Conclusions<\/h2>\n

Before applying artificial intelligence to their HS code assignment process, the company spent a lot of time on this, and the codes were more frequently incorrect.<\/p>\n

However, by using Artelnics\u2019 machine learning technology, this company managed to eliminate manual data entry, reduce human errors, and improve efficiency.<\/p>\n<\/section>\n