Machine Learning Applications in E-Commerce at Flipkart

Abstract

Machine learning (ML) has emerged as the technological backbone of modern e-commerce, enabling platforms to deliver hyper-personalised shopping experiences, optimise dynamic pricing, detect fraudulent transactions, forecast demand with precision, and automate supply chain decisions at scale. Flipkart, India's largest domestic e-commerce marketplace with over 500 million registered users and more than 150 million product listings, has deployed machine learning across virtually every dimension of its platform operations since its acquisition by Walmart in 2018. This study examines the application of machine learning algorithms at Flipkart across six functional domains: product recommendation systems, dynamic pricing engines, demand forecasting, fraud detection, customer sentiment analysis, and logistics optimisation. Primary data was gathered through structured interviews with 20 technology and data science professionals associated with Flipkart's Bengaluru and Hyderabad operations. Secondary data was sourced from Flipkart's published technology blogs, academic literature on e-commerce ML applications, and industry research from McKinsey and Gartner. Findings confirm that Flipkart's collaborative filtering recommendation engine accounts for 35% of total platform revenue, that ML-driven dynamic pricing processes over 5 million price adjustments daily, and that the fraud detection system achieves 97.3% accuracy with a false positive rate below 0.8%. Challenges include data privacy compliance under India's Digital Personal Data Protection Act (DPDPA) 2023, coldstart problem in recommendations for new users, and model explainability requirements for pricing decisions. Recommendations focus on federated learning for privacy-preserving personalisation, explainable AI frameworks for pricing transparency, and reinforcement learning for real-time inventory optimisation.