Beyond Surveys - Real-Time User Satisfaction Prediction

Learn how to revolutionize customer satisfaction measurement by moving beyond traditional surveys to real-time Net Promoter Score (NPS) prediction using machine learning and telemetry data in this 13-minute conference talk. Discover the fundamental concepts of NPS, including how it measures customer loyalty through detractors, passives, and promoters, and understand the four major limitations of conventional survey-based approaches that make them slow and ineffective. Explore the innovative Dynamic Net Promoter Score (DNPS) system that predicts customer sentiment without requiring surveys by leveraging behavioral data, telemetry, and text feedback through a sophisticated AI architecture. Master the complete data processing pipeline from ingestion to insights, including data cleaning, alignment, and feature extraction techniques. Examine the hybrid modeling approach that combines structured data models, sequence models, and text models through a fusion layer enhanced with explainable AI techniques like SHAP and LIME. Analyze real-world results demonstrating accuracy improvements, ablation studies, and real-time latency performance, while understanding how explainability insights drive business value by identifying key satisfaction drivers. Gain insights into the limitations of current approaches and future research directions for advancing real-time customer satisfaction prediction systems.