Data challenge
With hundreds of thousands of products and diverse data points, the company needed an efficient way to classify and manage their product data. Historically, this process had relied on manual input through Excel spreadsheets, which was slow, prone to errors, and unsustainable given the scale of their operations.
The manual nature of the process not only slowed down operations but also increased the likelihood of inaccuracies in product classification. Recognising this inefficiency, the organisation sought to streamline their processes. The goal was to find a solution that would reduce the labour involved in inputting and amending data while improving accuracy.
In addition to the data management challenges, their business model required a balance between standardisation and flexibility. Store managers needed autonomy over their individual operations, while adhering to company-wide product classification standards. This delicate balance added complexity, as the system had to be flexible enough to allow for individual store needs, while ensuring consistency across the organisation.
Solution: Product classification
Classifying products for Troy was particularly challenging due to the complexity and diversity of their offerings. The company required a system that could accommodate a four-level hierarchy, enabling products to be categorised with increasing levels of detail as needed.
To address this challenge, we collaborated with the organisation to develop a dynamic classification system. This solution involved building a flexible, four-level hierarchy that allowed products to be categorised at any level, depending on the level of detail required.
- Clothing (most general)
- Workwear
- Gloves
- Nylon Gloves (most detailed)
With over 30,000 level 4 categories, the system had to accommodate complex classifications while remaining adaptable. We developed a proprietary system that only allowed products to be quickly assigned to categories. Notably, while the system’s inherent language knowledge remained constant, it became better at understanding specific product categories, learning dynamically from previous assignments.
The system was designed to rapidly churn through 1,000 to 2,000 products per hour, providing suggestions for categorisation with up to 80% accuracy. The remaining 20% required further investigation and manual adjustments. Over the course of the project, the organisation used this system to process at least 18,000 products.
A key feature of our solution was a user interface that displayed the top five categorisation suggestions for each product. For more complex products, the system provided a deep dive into additional suggestions, enabling users to access more detailed product data, such as Google searches for product codes, which were run through the system’s engine to extract additional information. Importantly, the entire system was built with proprietary technology, ensuring complete privacy with no reliance on third-party APIs - a critical element that Troy valued.
Through a series of iterative builds, Amplifi worked with the organisation to continuously refine the input mechanism, incorporating feedback to improve both the accuracy and usability of the system.
Result: The project outcome, and Amplifi’s value
The implementation of the product classification system led to significant improvements in their data management processes. Manual input was drastically reduced, errors were minimised, and overall operational efficiency increased. As the system processed more data, it evolved, becoming progressively better at classifying new products, which further reduced the workload for employees.
Beyond streamlining the classification process, this project also opened the organisation’s eyes to the wider potential of AI. The foundations laid by Amplifi not only addressed their immediate challenges but also set the stage for future AI and machine learning (ML) initiatives. Since the project, they have expanded their AI capabilities, building on the principles introduced by Amplifi to develop their own machine learning infrastructure. This has allowed them to continue innovating and enhancing their operations.
Our role as an accelerator and scaffold was critical to the organisation’s success. By helping to build a self-sufficient AI capability, we empowered the engineering distributor group to tackle future challenges independently. Our approach was focused on solving real business challenges, not simply implementing AI for the sake of it. The Amplifi team provided a practical, results-driven solution that delivered immediate value and enabled long-term growth.
Our team’s expertise in foundational aspects of AI, combined with a focus on practical implementation, ensured the system wasn’t just a flashy, short-term solution but a sustainable capability that continues to deliver value.
Amplifi is committed to developing and supporting businesses as they build their own AI and ML capabilities. We act as an accelerator, providing the initial scaffolding needed to support growth. Once the foundational building blocks are in place, we step back, empowering organisations to continue their journey independently, ensuring they don’t become reliant on external support.
If you have an AI-based idea you want to turn into reality, read our guide for 6 expert tips for adding business value with AI below. Not sure where to begin? Our data experts can help you determine your needs and devise a strategic approach to building AI/ML capabilities, tailored to your organisation’s unique goals. Get in touch here.
