Ushering in a new era of personalized retail shopping

Overview

Emart24, a Korean convenience store chain with over 6,500 stores, migrated to the Databricks Data + AI Platform to process transaction data from more than 6 billion transactions and 80,000 products, cutting analysis time from 27 hours to 1 hour (96% reduction) and lowering compute costs by 93%. The new platform underpins AI-powered product recommendation services, a cashierless smart store system, and demand forecasting for logistics centers.

The challenge

Emart24's data was scattered across multiple systems, including sales systems, POS systems and customer applications, making it difficult to manage daily transactions and operational data from more than 6,500 stores and 13 large distribution centers. Their legacy data platform required a lot of time and high resource costs to analyze more than 6 billion transactions and data for over 80,000 products, taking 27 hours to run a single analysis at a resource cost of 930,000 won, making it virtually impossible to provide to store management daily.

The solution

Emart24 migrated to the Databricks Data + AI Platform on Azure, enabling data compression, skipping, partition pruning and caching, plus batch scheduling management, automated cluster management and task monitoring. The company built a data culture for cross-team collaboration and shared data with partner Shinsegae I&C via Delta Sharing for joint analysis. On the platform, Emart24 built AI-powered product recommendation solutions optimized for specific stores and locations, and pursued cashierless operations and improved demand forecasting.

Reported business value

Emart24 reduced analysis time from 27 hours to one hour, a 96% reduction that helped launch new products and features such as cashierless shopping. Autoscaling and optimized clusters reduced cloud spend from $930,000 to $70,000, a 93% reduction in cost.