There is an increasing demand for highly accurate positioning information in urban areas, to support applications such as people and vehicle tracking, real-time air quality detection and navigation. However systems such as GPS typically perform poorly in dense urban areas. A number of authors have made use of 3D city models to enhance accuracy, obtaining good results, but to date the influence of the quality of the 3D city model on these results has not been tested. This paper addresses the following question: how does the quality, and in particular the variation in height, level of generalization and completeness and currency of a 3D dataset, impact the results obtained for the preliminary calculations in a process known as Shadow Matching, which takes into account not only where satellite signals are visible on the street but also where they are predicted to be absent. We describe initial simulations to address this issue, examining the variation in elevation angle – i.e. the angle above which the satellite is visible, for three 3D city models in a test area in London, and note that even within one dataset using different available height values could cause a difference in elevation angle of up to 29°. Missing or extra buildings result in an elevation variation of around 85°. Variations such as these can significantly influence the predicted satellite visibility which will then not correspond to that experienced on the ground, reducing the accuracy of the resulting Shadow Matching process.