iSell
Industry
Location
Employees

Established in 1998, iSell is a global leader in IT lifecycle management. Their award winning Cloud-based software quoting platform ITQuoter helps MSPs and VARs automate quoting and procurement. They operate a web-based software service called IT Quoter, which contains a catalogue of over 5,000,000 products and allows companies to easily build and receive quotes on large quantities of IT products and software solutions.

The Challenge

IT Quoter has a catalogue of more than 5,000,000 products, with additional products regularly added in the fast-paced world of tech. In order to add these new products to their system, they must first be tagged and categorised into the correct product categories using information provided by the manufacturer. Unfortunately, due a lack of consistent industry standards, the information provided by the manufacturer is often incorrect or of poor quality.

In order to overcome the challenge of inconsistent data quality offered by the manufacturer, iSell has a team of product specialists that have to manually categorise these products into the right categories. Due to the complex nature of this task, some products may take up to 30 minutes to categorise. In an effort to speed up the process, a machine learning model was requested to augment iSell’s expert team to increase the speed of product categorisation. 

The Solution

Given iSell’s large catalogue of products, we developed a custom Machine Learning model to classify the products automatically based on features such as product name and description. 

The machine learning model was trained on over 1,000,000 products and hosted on the Google Cloud platform. This scalable architecture is capable of handling all types of product data. We also developed a custom data pipeline which makes use of big data and machine learning products offered by Google Cloud to manage the data-preprocessing, model training and deployment. 

This unique approach frees up iSell’s expert team to focus on the most complex products.

The Outcome

- The AI model has classified over 116,000 unlabelled products into over 1250 categories with comparable accuracy to human labelling. 

- The AI model is able to categorise millions of new products within minutes, saving hours of manual time and cost.

- The AI model can classify new products on-demand or as a scheduled batch job, and can constantly be updated with new product data. 

- The result of this AI model has helped iSell save hours of manual categorisation and allows their team to focus on the most difficult products to categorise. 

- As the AI model can continuously learn and improve with new products fed through the model, the accuracy of the product categorisation increases over time.