Devised future store strategy for a Russian retail chain using ML-based geospatial data analysis

Devised future store strategy for a Russian retail chain using ML-based geospatial data analysis

Devised future store strategy for a Russian retail chain using ML-based geospatial data analysis


Client overview

The client is one of the largest retail chains in Russia, with more than 200 hypermarkets having an average selling space of 6200 square meters. Founded in the 1990s, the retailer employs more than 35000 people and have an approximate total revenue of $5 billion.Challenges
The retailer wanted suggestions for the strategic positioning of their future stores. They recorded the geospatial data of the local population, which needed to be mapped to the population-catchment data for suggesting the next stores.

However, the locations of the existing stores were eating away customers of the other outlets of the retailer in the adjoining areas. Therefore, the data required an in-depth analysis to avoid cannibalism between other outlets of the retailer and competitors.

The retailer also maintained the data in silos, which made it difficult to use it across multiple use cases.

Solution
Algoscale developed a machine learning framework that considered factors like competitor stores, population density, houses/apartments, consumer pointer index (CPI), and nearby population catchments to strategize locations of the future stores.

The ML-based geospatial data analysis revealed that the aggregated financial status of a population cluster was directly impacting the sales of a nearby store. Algoscale combined the financial and catchment data to gather insights on how people with multiple economic backgrounds are traveling towards nearby stores. It also took liquor sales data into account, which was a significant revenue generator for the stores and was earlier ignored by the retailer.

The algorithms used in the solution gave valuable performance insights, thereby decreasing inter-store cannibalization, increasing new customers for the new stores, and reducing competitor’s sales.

The solution also introduced a new data recording protocol, which involved a data nomenclature system with constant QA and increased ML applications across multiple use cases.

Highlights

  • Heatmap of the target area with declining and prospective customers
  • 15% sales spike in new stores, despite the presence of competitor’s stores in nearby areas
  • New store locations based on population-mapped catchment data
  • Visual, intuitive user interface

Recent Posts

Subscribe to Newsletter

Stay updated with the blogs by subscribing to the newsletter