Unlocking the potential of data science in retail at Distriplus

“The initiative was launched as part of the DEL20 innovation ecosystem,” explains data scientist Kevin De Beck, one of the driving forces behind the project. “Distriplus was sitting on a lot of customer data due to its elaborate loyalty program. The original request was to find ways to valorize that treasure trove of information. But once we got started, it became much more than that.”

A two-phased approach

In the first phase, delaware and Distriplus dove into the client population data. “We wanted to know what the Distriplus client demographic looked like, and who their most loyal fans were,” explains Kevin. “Our hope was that this information could tell us why some customers stay with the brand for years, while others never return although they’ve subscribed to the loyalty program.”

In the second, more advanced phase, Distriplus and delaware wanted to zoom in on buying behavior within distinct product categories. This would allow the company to make personalized recommendations based on what customers had bought before. “For online web shops, that’s relatively simple,” says Kevin. “The only problem though: when the project started, Distriplus didn’t have an online store yet. So, the challenge for us was to find out how we could bring the project to brick and mortar stores.”

Onboarding and retention

One of the things Distriplus learned from its phase 1 analysis was that onboarding customers went a lot smoother than retaining them. To find out why, Kevin and his team proposed setting up a proof of concept in one of the group’s Di shops. In this pilot project, the team would take a closer look at the items individual consumers bought, how that related to their loyalty to the brand, and which items were most often bought together. The goal was to improve individual sales and retention rates for Di shops by uncovering cross-selling opportunities.

“In essence, we built an offline ‘recommendation engine’ for the shop,” says Kevin. “This was based on data gathered through the customer loyalty program. For example, we learned that certain items are often bought together, like toothpaste and mouthwash, while others are mainly individual purchases – like perfume. Through smart filtering, we managed to create a model that was applicable in-store.”

3% sales increase

In practice, this meant that the recommendation engine had to be translated into practical tips for store personnel. Kevin: “Based on the insights we gathered during the program, staff on the shop floor were trained on specific cross-selling opportunities. The result was almost instant: in just a few days, sales in the pilot shop increased by 3%.”

For Distriplus, the project was proof that data science isn’t just valuable for online retail, but could play a role in brick and mortar stores as well. At the time of writing, the group is in the process of extending the project to more Di stores across Belgium.