Are companies solving the wrong problem? While product teams race to launch new features, many overlook a crucial truth: users don’t leave apps because they lack features—they leave because the app doesn’t meet their specific needs.
Arbaz Khan, a veteran in building large-scale AI systems, asserts, “Companies need to compete on understanding users, not feature lists.” Today’s technology makes personalization not just possible but essential for success.
India’s Digital Revolution: Personalization at the Core
Khan highlights India’s food delivery giants, Swiggy and Zomato, as leaders in personalization. “A user doesn’t want to sift through thousands of restaurants—they want to see their favorite Biryani place or options matching their taste and budget,” he explains. Similarly, streaming platforms like Hotstar and Netflix thrive by catering to local preferences, such as offering regional content or personalized sports highlights.
Personalization Beats Features and Content
“Even the best content is useless if users can’t find it,” Khan emphasizes. In e-commerce, platforms like Flipkart and Amazon succeed by tailoring search results. For instance, a search for ‘kurta’ shows options based on past purchases, style preferences, and budget, which drives higher conversions than simply showing the highest-rated products.
Affordable Personalization: A Reality
Khan challenges the notion that personalization is costly. “A consumer-grade PC can handle 10,000 predictions per second with deep learning models, while gradient boosting systems can process up to 100,000 predictions on the same hardware,” he reveals. Personalization, he explains, can be both efficient and budget-friendly.
Smart Architecture: The Two-Stage Approach
Khan outlines a two-stage recommendation system that balances speed and quality. The first stage uses a lightweight model to shortlist 100 candidates quickly. The second stage employs a deep learning model to re-rank these for the user. This approach powers platforms like Flipkart and Uber Eats, enabling real-time, cost-efficient personalization.
The Future of Personalization
Investments in personalization technology are growing. “Top recommendation models today reach sizes of 100 terabytes, requiring significant infrastructure. However, the ROI is undeniable,” Khan states. Personalization directly impacts user engagement, retention, and revenue.
Khan concludes, “As India’s digital economy expands, the companies that succeed will be those that best understand their users. The future of digital products is personalized, and those who invest in this will lead the market.”
Arbaz Khan is a machine learning expert in recommendation systems. His views are personal.
