The dynamic vehicle routing problem (DVRP) presents significant logistical challenges, particularly in scenarios where service requests are subject to variations in both space and time. Accurate request forecasting is key to optimizing routing decisions and enhancing various performance metrics in DVRP. This research explores the impact of different request forecasting models on routing optimization, evaluating their effectiveness across various scenarios and strategies. By conducting extensive experiments, the study seeks to determine whether the choice of the best forecasting model leads to the best routing decisions, and examines the value of incorporating stochastic knowledge for improving DVRP outcomes. Spatial and temporal patterns of request arrivals, including uniform and normally distributed spatial patterns and constant arrival rate, sinusoidal rate with one bump, and sinusoidal rate with two bumps, are investigated. While the primary objective is to minimize unserved requests and travel time, the research also explores additional metrics such as response time to assess the influence of forecasting on routing solutions.