Understanding customer purchasing behavior in shopping malls is essential for developing marketing strategies to boost overall sales. Traditional customer flow analyses, through measurements or simulations, have not sufficiently accounted for behavioral changes due to time or events, failing to capture the variations in customer behavior across different situations. This paper proposes a method that dynamically captures changes in customer behavior by sequentially incorporating measurement data into simulations. We employ a particle filter, a data assimilation technique, to agent-based simulation to dynamically estimate the customer’s transition tendencies between stores. The evaluation of the proposed method involves estimating the transition tendencies in a virtual shopping mall based on real-time inflow counts of customers. The experiments show that the proposed method can qualitatively estimate the dynamically changing customer’s transition tendencies. Implementing this approach could afford a dynamic insight into customer behavior, aiding in the formulation of effective strategies to drive sales and improve safety.