In this paper, a multi-step model related to freight movements in urban areas is formalised to solve a Vehicle Routing Problem with Time Windows (VRPTW). In modelling terms, the VRPTW is formulated to consider the optimum paths between all the customers combined to determine the best vehicle routes. The optimum path search with a selective mono-criterion approach is tackled. A model for forecasting the link cost is calibrated. Some procedures (traffic assignment, real-time system measurement, reverse assignment) for estimating system performance in terms of travel cost are also proposed. In terms of procedures, the VRPTW is solved with a genetic algorithm and two different crossover operators are reported and used simultaneously. A real case application is made, comparing the results of the proposed procedure with the routes of a truck (<6 tonnes) delivering dairy products to some retailers in a city. The contribution of this paper is to relate the Vehicle Routing Problem (VRP) to transport the network theory. Indeed, the VRP is generally treated without considering network congestion. In the approach proposed in this paper, the costs are assumed to be functions of network characteristics and traffic flow, while a model for forecasting link costs is calibrated from real data.