Selection of appropriate web services is an important step in development of composite applications. Quality of Service (QoS) data characterizing nonfunctional properties of candidate web services are usually used in web service selection. However, functional characteristics, which are difficult to measure, are equally important. Quantitative evaluation of functional characteristics is possible in the case of development of composite decision-making applications. This paper elaborates a web service selection method, which accounts for both functional and nonfunctional characteristics of candidate web services. The functional characteristics are evaluated according to decision-making results yielded by the composite application. A linear programming model is used as the web service selection method and the cost of developing and operating the composite application is used as a selection metric. Application of the proposed web service selection method is demonstrated by developing a composite application for a vehicle routing problem, where web services are required to provide spatial data necessary for the decision-making algorithm. Experimental results show that accounting for functional characteristics substantially affects web service selection results.