Real-time Data Processing and Analytics for a Global Retailer
Overview.
A global retailer was able to harness real-time analytics to process and analyze a staggering 100,000 sensor messages per second from their IoT network. This enabled them to predict maintenance needs for cold storage facilities and transportation vehicles. This proactive approach resulted in a significant ~10% reduction in food waste and a ~15% increase in the shelf life of perishable items.
Our client, a multinational retailer operates discounted stores and hypermarkets, and installed IoT sensors to monitor food storage, transport vehicles, cooking equipment etc. for the whole year round. These IoT sensors sent approximately 100,000 messages per second, which were to be parsed and enriched in real-time. Data science teams were looking to process these enriched messages to ML models to ensure predictive maintenance of the IoT sensors.
Solution.
Coforge developed a real-time data processing layer using Kafka, Spark, Cassandra, Redis.
Implemented a data enrichment exercise to add master data attributes, tagging, business rules-based data quality checks.
Real time Data aggregation to build aggregate metrics, threshold checks, and live dashboard
Developed an ML model, for near real-time data analytics
Implemented Cassandra for data storage and analysis of historical data
The impact.
Seamless data ingestion & processing generated real-time predictive insights Architecture for high performance & scalability Predictive maintenance of sensors resulting in ensuring of quality perishable items & food safety Quick decision-making through interactive dashboards
