Intelligent UI Rendering: Leveraging Machine Learning to Predict and Preload User Journey Steps in E-Commerce Sales Funnels

Abstract

Page load speed has a direct impact on whether shoppers complete a purchase or abandon the funnel. Despite years of investment in optimization techniques like code splitting, lazy loading, and CDN delivery, most front-end applications still treat every navigation event the same way. They do not account for what a specific user is likely to do next. Existing tools like Guess.js and Next.js do use historical data to prefetch common routes, but they rely on averaged patterns across many users rather than signals from the current session. This paper describes the Intelligent Preload Engine (IPE), a lightweight machine learning framework that watches what a user is doing right now - how long they dwell, how far they scroll, how often they click - and uses that to predict their next navigation step before they take it. The IPE then preloads the relevant route bundle and API data in the background, so the transition feels instant. In simulation across a five-stage e-commerce funnel, the IPE achieved next-step prediction accuracy of 76.8% with a Markov model, 81.3% with an LSTM, and 83.7% with a lightweight Transformer. Simulated LCP on predicted targets dropped by an average of 58%, and funnel completion was projected to improve by 2.1 percentage points. These results come from simulation and will need to be validated against real production traffic before drawing firm conclusions.