Case Beslist.nl

Beslist.nl is a Dutch website which allows its visitors to compare consumer products. The website is not a store where you directly buy something: when you click on a product, you will be redirected towards the seller. With an immense product range, the brand faced some difficulties in categorizing products.

Challenge

Beslist.nl provides over 36 million products of 10.000 shops on its website, which makes it an immense chaos: products are not always in the right (sub) category, which complicates the process of filtering products. So, in order to make their website clearer and to improve the customer experience, the brand looked for solutions that allowed them to efficiently categorize products. Manually categorizing all of these 36 million products would take an immense number of manhours, making it unfeasible and way too expensive. That is why the brand looked for solutions to automate this process. This is where Squadra Machine Learning Company offered to help find a solution for this issue.

Solution

Artificial Intelligence is a relatively new area of techniques where innovations evolve rapidly, and its appliances are broadly extended. Machine Learning Company can assist Beslist.nl with multiple services, one of which is product classification. In order to receive a more efficient classification process, Squadra MLC has created subcategories for the already existing categories. Within these subcategories, filters were created so that customers were able to filter in even more depth. After that, Squadra used Machine Learning to predict in which category a certain product belongs.

Machine Learning is computer science that gives computers the ability to learn without being explicitly programmed. Instead of writing lines of code with a lot of rules like “if something then do this”, Machine Learning software identifies patterns and creates algorithms based on large sets of data. It is an extremely powerful technique which is able to perform intelligent tasks which were up to now usually done by humans. It can recognize, predict, advice, optimize and classify, and therefore it can support the automation of various business and industrial processes, customer facing services and other data driven functions.

To be sure that the method Squadra developed was effective, a proof-of-concept was executed. This resulted in an accuracy of 89-94%, which led to the conclusion that the algorithms worked properly. After this, all of the remaining products were inserted into the smart tool, but it appeared that the algorithms could be improved for certain categories like electronics and household products. In order to achieve an even better result, Squadra’s team is still developing new algorithms to classify these categories.

Result

After collaborating with Squadra, Beslist.nl was able to categorize their products in a more efficient manner. With over 36 million products, manually classifying would be an infeasible process that would cost way too much time and money. The collaboration has not ended yet: Squadra’s team is still working on improving their algorithms so that the accuracy of 95% or higher can be achieved.