In this paper, we study the vehicle routing problem with dynamic customers, where a portion of the customer requests are known in advance and the rest arrive in real time. We propose an optimization-based look-ahead dynamic routing framework that involves request forecasting, partial planning, and dynamic real-time routing of the fleet. This framework has the capabilities for adjustments in response to routing environments with different levels of uncertainties. Through extensive numeral simulations, we exam its performance in routing environments with various levels of uncertainties. We demonstrate the efficiency and robustness of the proposed solution by benchmarking against two other routing strategies. This paper fills the gap in the literature on studying the relationship between the level of route planning in the solution approach and the quality of the solution under various system conditions.