Building a Recommender Engine for a B2B Webshop

Adomex International is a specialist in the import and export of all types of dendrological and herbaceous decoration greens. With five locations throughout the Netherlands, Adomex is respected, not just as a trusted supplier within the industry, but as a collaborator and innovator working to connect customers with sustainable and equitable sources and harvests for floral and decorative assortments and arrangements. Their customer base embraces wholesale to large export companies, to flower shops along the Dutch canals and abroad.

Improving a Green Network

Adomex has become more dependent on a consistent data system with the demand, quantity and diversity of products and transactions increasing every day. They turned to Aurai to activate their daily quantitative knowledge into immediate practical improvements derived from engineering and automating information.

A Recommender Approach

First, our consultant interviewed stakeholders across the organization to assess the culture, context and organizational structure in order to match the most pertinent information sources with the best available applications. There are many potential gains from a variety of data analytics, but time and energy are maximized when the engineering is targeted and specific. Together with these stakeholders we came up with a project designed to make a direct impact on Adomex’s revenue generation. We went beyond a classical data science project to empower a company to make data-driven decisions and efficient customer pathways through recommendations. 

A Sales-Oriented Solution

Bob, one of Aurai’s data scientists, began by building the first version of the recommender engine alongside Adomex. Bob divided the project into multiple sprints, and after each sprint, Bob presented its progress and discussed any obstacles and the best strategy for moving forward. We involved Adomex at every benchmark to ensure the model was perfectly tailored for our client. 

The recommender is comprised of multiple modules: candidate generation, which is subdivided by a content-based filter and a collaborative filter; and filters for candidate ranking and stock availability. The parameters and weighing factor of the filters were configured to clarify and simplify, thereby increase, Adomex’s sale processes. 

A Smarter Webshop

After reviewing and validating the program, the recommender proved to be performing very well. Adomex has implemented the model into its webshop where it is currently running, bringing the forest and fields to customers’ homes and generating revenue for a competitive ecologically sustainable business.