Improving Personalized Product Recommendations in Flipkart's Shopping App

You are a product manager at Flipkart, the Indian e-commerce marketplace. How would you improve its AI-powered product recommendations? Explain which shoppers and recommendation problems you would focus on, what you would build, how you would prioritize it, and how you would measure success. ```hint Look for where recommendations fail Walk a shopper's journey from browsing to after delivery and note the moments when a suggestion is irrelevant, repetitive or badly timed. Those failures point to better improvements than simply swapping in a newer model. ``` ### Clarifying Questions - Is the goal more purchases, larger orders, better discovery for sellers, or long-term retention? - Which recommendation surfaces are in scope: the home feed, the similar-item and bought-together widgets on product pages, the cart, or notifications? - Should I focus on buyers only, or also on how recommendations affect sellers and sponsored listings? - Are any categories or shopper groups a priority, such as first-time online shoppers from smaller cities or people using regional languages? ### What a Strong Answer Covers - A clear goal for recommendations tied to Flipkart's business, and a focused shopper segment with reasons - Concrete pain points in today's recommendations, located on the shopping journey - Solutions that apply AI where it fits (intent, context, compatibility, natural-language requests) and a prioritization by impact and effort - Success metrics that capture incremental purchases rather than clicks, with guardrails such as returns and seller fairness - Risks such as privacy and consent, cold start, sponsored content and load during big sale events ### Follow-up Questions - Clicks on recommendations rise but total orders do not. How do you investigate? - How would you recommend products to a first-time shopper with no history? - Where should sponsored products sit in a personalized feed without eroding trust?