Data analysts play a crucial role in organizations by transforming raw data into actionable insights. This study examines how their productivity evolves on a collaborative platform through two key learning activities: writing queries and viewing peer queries. Traditional static models fail to capture the dynamic nature of learning, so we propose a Hidden Markov Model (HMM) to track analysts’ transitions between novice, intermediate, and advanced learning states. Analyzing 79,797 queries from 2,001 analysts, we find that productivity improves with state progression, though moving from intermediate to advanced state is particularly challenging. Writing queries consistently improves learning, while viewing peer queries helps novices but can hinder experienced analysts due to cognitive overload. These findings provide insights into dynamic learning behaviors among knowledge workers and offer practical implications for optimizing training, enabling personalized learning, and fostering effective knowledge sharing in designing effective collaborative systems.