🚨 So how do you predict which ones will become high-value customers — before they've even placed one order? Swiggy's answer wasn't a bigger AI model. It was 350+ razor-sharp features (GPS spoof detection, phonetic name-matching between neighbors, 4-layer geo-density) feeding a surprisingly simple neural net. The twist: adding a second prediction task shrank the model by 63% — and made it more accurate. Then they beat a paid third-party platform in a live A/B test. At zero extra cost, get the details 👇 https://lnkd.in/g69Z5b2s Jayshmi A Sunil Rathee #MachineLearning #DataScience #Swiggy #Instamart #QuickCommerce #AI #FeatureEngineering #GrowthMarketing #Ecommerce #lifetimevalue
You are running an e-commerce platform. You want to improve retention. You want to personalize discounts. You want to optimize marketing spend. You want to catch fraud before it happens. Every one of these decisions depends on a single question: ❓How much is this customer worth? That’s where predictive Customer Lifetime Value comes in. I recently got the opportunity to write about how we think about cracking this problem at Swiggy - from the intuition behind pLTV to the modeling ideas and practical challenges of estimating customer value at scale. I hope you enjoy the read. https://lnkd.in/g69Z5b2s
