This Is How HS Code Classification Can Be Smart

HS classification in Singapore

The ASEAN Harmonised Tariff Nomenclature (AHTN) consists of a Harmonized System (HS) structure up to the 8th digit. Singapore has adopted this HS structure.

The General Interpretative Rules (GIR) is a widely adopted guideline of HS Classification. It is time-consuming and challenging to manually identify the appropriate HS code for each new product or when a product description / specification changes. It requires the classification to base on the material composition of the product.

As such, a change of percentage in material composition could result in an HS code in an entirely different chapter.

The issue is exacerbated by fines imposed by Singapore Customs for misclassification of HS code which can be as high as SGD 10,000 according to SG Custom's circular SG Custom's circular:

"A person guilty of making an incorrect declaration under the Customs Act is liable on conviction to a fine up to S$10,000 or one time the amount of duty and GST payable, whichever is higher, and/or imprisonment for up to 12 months."

Companies are offering professional services to classify HS code for the product(s). However, in industries where new products are frequently imported and products' description / specification is fast-changing, this will not be an economically viable option. Examples are Fast-moving consumer goods (FMCG) and e-Commerce.

New IT gadgets are fast evolving, hence have complex and/or innovative combinations of description due to their new features.

Consider a real-life example using the product description of a smartwatch posted on an eCommerce site.

The product has a description of "Smart Watch Sport Band Bluetooth Jam Pintar Smart Watch Wristband Fitness Tracker Heart Rate Monitor Men Women Y68".

Other HS classification systems may classify this product description into incorrect chapters of the HS nomenclature.

For example, the following screenshot shows the HS classification system lists chapter 61 and 62 (both covers "Articles of apparel and clothing accessories") as the top results for the above product description of a smartwatch:

Classification into incorrect chapters by a HS classification system
Classification into incorrect chapters by a HS classification system

This happens because the system is using a simple keyword matching algorithm. So, when the system finds the word "women" in the product description matching the same word in HS Chapter 61's name, it mistakenly lists Chapter 61 for this product description.

In contrast, GeTS's CALISTA Intelligent Advisory (CIA) understands that this product description is referring to a smartwatch. It shows the correct HS chapter, heading, and subheading (refer to the following screenshot).

SG HS for "Smart Watch Sport Band Bluetooth Jam Pintar Smart Watch Wristband Fitness 
                        Tracker Heart Rate Monitor Men Women Y68"
SG HS for "Smart Watch Sport Band Bluetooth Jam Pintar Smart Watch Wristband Fitness Tracker Heart Rate Monitor Men Women Y68"

Since HS code is harmonized at the 6-digit level among 200 World Customs Organisation (WCO) members, this ruling by United States Customs suffices to prove that 851762 is generally accepted by customs worldwide as HS subheading for smartwatches.

The smartwatches can have vastly different product descriptions due to a variety of technologies, functions, and purposes. For example, some low-end smartwatches might not have Bluetooth features; some smartwatches are unisex, but some others are designed for men or women exclusively.

So, do you want to pay a hundred dollars to a professional service company to classify each item you import that is of a different description?

The Solution

Let us show how to smartly classify many products into the appropriate HS code.

Advances in Artificial Intelligence (AI) and Machine Learning (ML) thereof have helped propelled progress in text analytics and aided in Natural Language Processing (NLP) of layman product description.

GeTS has developed AI models that receive input of product description, process it through the natural language process to identify the more appropriate HS Code to be recommended.

This process considers the statistics of past permit declaration to Singapore Customs, and other proprietary algorithms.

ML requires quality data to train the model to better recognize and understand the text description of products. This is one major aspect where GeTS is strong.

GeTS has been a Singapore Customs-accredited front-end provider for import permit declaration for many years. The accumulated data add value to GeTS HS Classification engine. It does so by training the models in the engine to recommend HS codes that are matched or closer to the HS codes approved by Singapore Customs in the past. It is done based on the provided product description and other helpful information of the product as stipulated by GIR.

Enterprise-ready APIs for seamless integration

As part of the mission to power global trade connectivity, GeTS offers AI-powered CALISTA Intelligent Advisory (CIA) which includes an HS classification feature for various communities of users.

This service can be easily integrated via Application Programming Interface (API) with internal systems of companies who need to classify their products into HS code in an automatic manner.

For example, eCommerce platforms that support cross-border deliveries need HS code of the products, so that they can use the HS code to find out the estimated duties and taxes imposed by destination countries.

These platform operators can easily retrieve HS code for each product to be classified by programming their platforms to call the API of CIA with product description. Preferably, the description shall be as detailed as possible so that the recommended HS code would best represent the product. This integration saves a tremendous amount of operating expenditure, compared to maintaining a trade compliance operation. Manual operation is not only subject to human error during the classification, but also additional effort in keeping their compliance knowledge and product knowledge up to date.

How smart is your HS classification?

Many "HS classification" or "HS search" service or websites in the market are using simple text searching, in which, the product description entered has to be matching part, if not whole, of the official description in HS nomenclature, for the website to return a correct HS code.

For example, if you want to find out the Singapore HS code for "chicken satay", which is a local delicacy in ASEAN, many "HS search" websites expect you to enter texts matching the official HS nomenclature, such as "prepared or preserved meat"; otherwise, those websites might return HS code that is for a totally different product.

Wouldn't it be nice, that you simply enter "chicken satay" into the system, then the system returns you the HS code matching "prepared or preserved meat"? The good news is this is exactly what CIA is capable of!

Map from MY HS to SG HS for "chicken satay" using CIA
Map from MY HS to SG HS for “chicken satay” using CIA

Now, let's take a step further.

Our customs are the ultimate authority in ruling which HS code is for the product you import.

Wouldn't it be better, if the system also tells you how many past declarations containing the product description you enter and using that HS codes were approved by Customs?

For example, in the following image showing GeTS' HS search system, I entered "CHICKEN SATAY", and the system tells me that 2580 past declarations were using "CHICKEN SATAY" and HS code 16023290.

SG HS for "chicken satay" using GeTS TradeWeb HS Classification
SG HS for "chicken satay" using GeTS TradeWeb HS Classification

This is achieved by utilizing ML and NLP and takes historical commodity descriptions and HS codes that were previously approved by Singapore Customs as the major training source. The data then undergoes preparation and modeling before it is deployed. This is illustrated in the flow diagram below:

ML pipeline of HS Classification Engine in CIA and TradeWeb
ML pipeline of HS Classification Engine in CIA and TradeWeb

A sample of NLP processing is illustrated in the following screenshot.

The product description given below corresponds to a smartwatch catalogued on an eCommerce marketplace. NLP categorizes it into several possible categories with a confidence score. NLP is more confident that the given description corresponds to a product that more likely falls in "consumer electronics" than "tv and video equipment".

"Smart Watch Sport Band Bluetooth Jam pintar Smart Watch Wristband Fitness Tracker Heart Rate Monitor Men Women Y68"

Categorization of products using NLP
Categorization of products using NLP

Systems that leverage on NLP such as CIA then recommend HS codes that correspond to consumer electronics such as a smartwatch.

In contrast, many HS search systems classify this product description into apparel due to the appearance of the words "Men" and "Women".

This combination of big data and ML has yielded the following results in CIA:

Map from Chile HS to SG HS for "Miguel torres escaleras empedrado pinot noir" using CIA
Map from Chile HS to SG HS for "Miguel torres escaleras empedrado pinot noir" using CIA

The system provides users with time and costs savings, not only for the additional manpower required but also for potentially incorrect tariffs or fines. This in turn leads to faster clearance of declarations and approval from governing bodies.

Conclusion

You have learned a few things about HS classification:

HS classification is the basis on which many cross-border trade activities rely upon. These include but are not limited to qualifying your products to recap the benefits of lower preferential tariff provisioned by free trade agreements.

Consistently hitting the roadblocks in finding the suitable HS code could result in a thinning profit margin especially in cost-sensitive industries such as logistics that are already facing cutthroat competition amidst the COVID-19 pandemic.

CIA Quick Search! comes to rescue with an AI-powered look-up tool that supports both HS code and laymen description searches. It also provides up-to-date tariffs for over 180 countries with preferential tariff advisory. Get instant access with 10 free searches today!

Get an Instant Trial Account Now!