Blog
January 19, 2024
Companies that supply product data to their customers or receive data from various sources are familiar with the problem: every company has its own unique data structure. Classifications have often evolved over time, reflect the company’s way of thinking, or are highly granular. However, structures that make sense for a particular company can pose major obstacles to data exchange. Different product group structures come together, but they must be merged into a single, comprehensible system in the target system, such as a web shop or an electronic catalog.
Classifications – when data speaks a common language
In e-commerce in particular, the volume of data and the requirements of customers, buyers, and retailers make it crucial for products to be presented in very clear structures. After all, in today’s fast-paced (working) world, products need to be quick and easy to find, filter, and compare. Standardized data structures also offer great advantages to those who work with data. This is because if the product and catalog database is the same, formatting and adjustments are not necessary, and there are standards that all parties involved can use as a guide, this reduces the high workload involved in product data management.
In order to meet these requirements for common structures, a large number of product classifications have been developed in recent years. These can be maintained in addition to the company’s own data structure and classify products according to a system based on characteristics. In most cases, the classifications are hierarchical, so that products can be clearly found by all users across multiple levels. Together with clear descriptions, keywords, and explanations, goods from different manufacturers can be quickly and easily identified in this standardized system.
Both suppliers and customers benefit from this standardized classification of all items. For example, purchasing is made easier because the portfolio can be bundled according to consistent criteria, which facilitates item comparison in daily business or in tenders. Products can be easily found and identified, which ensures significantly greater speed and reliability, especially in complex processes such as delivery or processing. The clear description also increases the visibility of items on marketplaces and facilitates integration into shop systems, which strengthens sales and increases customer satisfaction. Three classification systems in particular have established themselves on the market: the UNSPSC classification from the USA, the European ETIM system, and the ECLASS classification established in Germany.
UNSPSC, ETIM, and ECLASS—the three major players and their distinctive features
The most comprehensive of these three classifications is the United Nations Standard Products and Services Code, or UNSPSC for short. This code can be used to classify both goods and services. It is used and required internationally, but is particularly prevalent in the US. UNSPSC-classified items are assigned an 8- to 10-digit code consisting of 4 to 5 descriptive hierarchical levels. Use of the code is free, but the comprehensive, annually updated complete code sets with around 157,000 different item classes are only available to members, which makes it difficult to classify one’s own products.
The European Technical Information Model (ETIM) is another system that was originally designed specifically for electrical engineering but is increasingly being extended to other product classes. Unlike UNSPCS, the 5,500 product groups are organized without a hierarchy. Although the classes are roughly grouped by topic, the lack of a hierarchy that could guide the user from the general to the specific makes it difficult to manually assign the 8-digit codes, especially when dealing with large amounts of data. Items in the classes can be described more specifically using attributes. The standard can be used without license fees and is regularly expanded.
ECLASS is another comprehensive standard that has established itself on the market, combining the features of the other two systems. Like UNSPSC, it uses a hierarchical classification system with four levels, so that items are described with an eight-digit code. Similar to ETIM, items can be assigned attributes, known as unique characteristics. With around 45,000 product classes and 19,000 characteristics, as well as new releases every year, ECLASS offers comprehensive classification options. Its use is subject to licensing, and the division into new classes or subclasses in particular poses major challenges for users when new releases are issued.
When the standard becomes a challenge
A look at the systems shows that classifications offer a solid and general standard for e-commerce. They simplify data exchange and use in different systems. However, a standardized classification system with its high requirements, special hierarchies, and annual changes leads to additional maintenance work. Since company structures rarely correspond to standardized classifications and cannot be transferred directly, items are usually classified manually. This monotonous task, which is performed in PIM systems or Excel, for example, requires a great deal of time and therefore human resources.
Even though service providers can assist with classification, correct classification often requires a high level of product expertise that is not always available. This results in lengthy queries or items being classified differently by different people, leading to misunderstandings and inaccurate data. In addition, products often have to be classified in several versions of the standard, which makes it necessary to maintain several data fields in parallel. Mapping tables are rarely freely available, and serving all standards is extremely complex due to the volume of data and the multitude of hierarchies and requirements.
The potential of AI-supported classification
Onedot has developed automatic product classification to meet these requirements and challenges. With the help of AI, large amounts of product data can be assigned to the appropriate hierarchies and product codes. The final check is carried out by the users, which improves the assignment with each use. The machine learning model was trained with classification data from public sources, so that the AI is familiar with the most common sales channels and marketplaces right from the start. However, the high potential of machine learning only unfolds when the amount of high-quality learning data increases. For this reason, Onedot offers you the opportunity to benefit from the community with opt-in product classification.
The classification data of the entire Onedot community is anonymized and made available to the AI for internal training purposes. This results in highly accurate and high-quality suggestions. If you do not wish to use this option, you can continue to receive classification suggestions based on public data via our opt-out option. However, with the opt-in option, you as a member will benefit directly and in the future from the community-supported, more accurate automated classification, so that you too are excellently positioned in the field of standardized product data.
Would you like to benefit from our platform and community in order to respond to market requirements in an agile and reliable manner? Then please feel free to contact us.