top of page

GenAI to automate product category classifications

  • Growing assortment sizes present a challenge to the traditional approach of manually categorizing products 

  • Generative AI in the form of Large Language Models like ChatGPT can help improve the speed and quality of classifying products 

  • 94% accuracy of product categorization was achieved, along with years of manual labor saved 

  • High data quality and continuous improvement are vital for obtaining the best possible results; automated self-assessment and quality control were added to the model  

At a glance:

Manual categorization is overwhelmed by cost and volume 

Efficient product categorization is crucial for assortment management, especially for e-commerce and wholesalers. However, relying on product experts to manually classify categorizations is becoming increasingly difficult as assortment sizes grow rapidly. Imagine the labor-intensive process of manually classifying thousands of products across various levels. Each classification decision involves diligent analysis, often resulting in significant expenses and delays.


A leading wholesaler faced this challenge: Assortment managers were overwhelmed by the sheer volume of over 200,000 products requiring classification. The manual process was not only expensive but also limited in scalability. With an FTE (full-time equivalent) capable of classifying just 20,000 products per year, the costs of wages alone would amount to more than € 600,000. Furthermore, manual categorization’s lead time spanned multiple years. The wholesaler needed a solution that would streamline the product categorization process, reduce costs and accelerate decision-making. Enter the GenAI project. 

Assortment managers and ChatGPT create synergy 

Our team embarked on an innovative project to automate product category classifications using the power of ChatGPT. We worked closely together with assortment managers to validate outcomes and to ensure the usability of the tooling and therefore also the solution. Here’s how it works:  

 

1. Collect high-quality data 

Assortment managers provided us with a list of product names and product numbers. Then GenAI used this to perform an online Google search, in which it collected Google screenshots of reliable sources, ensuring high-quality data extraction.  

 

2. Categorize products using ChatGPT 

The collected data was then presented to a Large Language Model (LLM), ChatGPT, which ‘understands’ what kind of product it is. The LLM followed the product tree structure provided by our client to determine the appropriate categories for each product.  

 

3. Automated self-assessment and quality control   

To enhance reliability, the GenAI tool self-assessed its results. High-quality results are directly implemented, whereas other results are used as a suggestion to the account managers, to improve the speed and quality of classifying the remaining products.  

 

Intelligent automation with GenAI: accurate, fast and flexible 

Within six weeks, the GenAI project achieved remarkable outcomes. The accuracy and confidence were very high as the LLM-based approach consistently delivered accurate category assignments, reaching up to 94% accuracy. Assortment managers manually verified samples to enhance trust in the AI solution and ensure quality standards.  

 

Faster categorization also translated into cost savings. Our client achieved substantial savings by automating the process. Additionally, it optimized lead time, saving years compared to manual work. Flexibility was achieved as product classifications can be swiftly adjusted when constructing a new category tree, allowing for flexibility.  

 

This GenAI project exemplifies the synergy between cutting-edge AI and practical business needs. As we continue to refine and expand its capabilities, we look forward to empowering our clients with intelligent automation. This GenAI tool isn’t just a success story for our client but a game-changer for similar use cases across industries:  

 

Retailers: Streamline product categorization, improve efficiency and reduce costs.  

 

Mergers & Acquisitions: Quickly match the product assortment of a potential merger to their product tree and visualize the overlap in the assortment.  

 

Repetitive Google Search Decision-Making: Automate processes that require a repetitive iteration of using Google search information to make informed decisions.  



Read more retail client cases


Ready to transform your business?

Stay ahead of your competitors by transforming your business. We have more than 30 years experience in helping clients innovate, deliver superior experiences and execute on their strategies. Curious how we can help?

bottom of page