fbpx

Code: EDI-2020-14-MIGROS

Domain: Retail

Summary

Predicting categorical sales in different stores accurately brings advantages in terms of operational efficiency and tracking the performances of stores or operational processes.

Proposed by

 

Migros is one of the largest FMCG retailers in Turkey. With more than 2000 stores and 30.000 employees, Migros is also the pioneer of organized retailing in Turkey. Migros today offers spacious stores in a wide range of formats and locations whose vast selection of cosmetics, stationery, glass and kitchenware, electronic appliances, book, textiles, and other items along with groceries and other necessities give it the ability to satisfy the shopping needs of its customers.

The company aims to be always the first choice of customers by providing a unique convenience and trustworthy shopping experience through its ultimate service approach, pioneer applications, broad product portfolio and family budget friendly pricing strategy.

Description

Migros is selling products in more than 100 categories in over 2000 stores. Predicting categorical sales in different stores accurately brings advantages in terms of operational efficiency and tracking the performances of stores or operational processes. Currently, the analytics team is building models based on product categories and stores, resulting in a huge number of models. Our objective is to decrease the computational requirements as well as increasing the performance of the predictions.

Data

The challenge has the following sample datasets available for download

Expected outcomes

Currently the sales are predicted with 5-7% accuracy (in total and in terms of mean absolute percentage error) by using models for each store and product category. Our objective is to increase the overall accuracy as well as the accuracy of store-category-based individual forecasts.

We also welcome any approaches that can allow us to reduce the number of models to be trained by keeping the accuracy roughly the same. We aim to reduce the number of models by 50%.

KPIs:

  • Exceeding 5% accuracy.
  • Increasing the accuracy of store-category-based predictions by 10%.
  • Decreasing the number of models by 50%.

How do we apply?

Follow by Email
Facebook
Twitter
YouTube
SOCIALICON