Machine Learning

Improving INTERSPORT's Product Assortment with Store Segmentation

INTERSPORT is on a strategic journey to fully leverage its data for decision making. As the first step of this journey, Datatonic has developed a store segmentation with machine learning based on the top product categories. As a result, INTERSPORT now has store segments to provide improved product assortment recommendations to its store network. The value of doing so is three-fold: offering an ideal product mix, simplifying the supply chain and improving negotiation with suppliers.

Impact
  • Developed a strategic store segmentation using Machine Learning on product mix per store
  • Enabled effective product assortment to lead to potential increase in revenue by improved coverage of customer needs
  • Enabled effective product assortment to lead to potential lower costs due to reduced time and complexity of decision making and supply chain
Challenge

INTERSPORT has over 5400 stores across 57 countries. These stores vary greatly in size, service level, product offering, and sometimes the banner they carry. Furthermore, customer demographics will vary as will local sports culture, all of which will affect general retail behaviour. 

Additionally, INTERSPORT is not a typical verticalized retailer, meaning store managers are given relative freedom of choice in terms of products for their stores. This situation makes any centrally steered initiative (e.g. product assortment allocation, communication) very complex to implement.

The scope of this project was focused on understanding if there are different meaningful types/categories of stores for which different strategies can be developed. Beyond this immediate scope, this project is part of INTERSPORT’s journey to leverage data to better understand and optimize its business.

“Datatonic made great strides in a very short time to understand the complexities of our business. Most impressive though is how Datatonic turned around and not only gave us a fresh perspective on our own numbers but also came up with comprehensive store types we can now act on.”

– Jan Haueter, Director of Business Development – INTERSPORT Group

Our Solution

Our goal was to segment the stores across all geographies, creating groups of stores with a similar mix of best-selling categories of products. This would provide INTERSPORT with the ability to develop bespoke sales strategies on the segment level rather than strategies based on geographies or sales volume.

The client data was securely moved into Google Cloud Platform. The data was uploaded to BigQuery, which allowed us to efficiently explore and transform the raw data. Google Cloud AI Platform Notebooks were then used to easily access a python environment and develop the clustering model along with pipelines to automate the process. A hierarchical clustering model was built to create the store segments. The segments were created based on the revenue share for each category of product for each store. Each category segment was additionally broken down into three sub-segments based on the total revenue of the store, to capture the “size” of the stores. This approach allows INTERSPORT to develop strategies on the segment level, based on the best categories of products to recommend to store managers. 

In addition to the main goal of the project, Datatonic identified additional external data for INTERSPORT to enhance their understanding of individual stores, including:

  1. Google Maps Place Data, providing information on the competitive landscape in a given radius and the ratings of each store using API calls
  2. ArcGIS Geo-Demographic Data, providing demographic and lifestyle information about a location or an area.

The store segmentation improved INTERSPORT’s ability to make internal and external category management decisions. This will reduce time to market and improve the relevance of the store assortment, ultimately leading to increased sales.