The pressure distribution generated over a building by wind is determined principally by the shape of the building. For basic building forms there are but four primary shape parameters - height, span, length and roof pitch - yet understanding the influence that each of these parameters has on wind loads remains suprisingly elusive and, consequently, design codes are unable to account properly for the effect of building geometry. Following recent refinements in other aspects of the wind loading design process, the prevailing crudeness of pressure coefficient data now seems to constitute the overriding limitation to the reliability of wind loading design. A programme of full-scale measurements on low-rise portal framed buildings has been completed and the results presented here provide reference points for the derivation of a detailed design procedure. However, since the full-scale buildings were selected from an existing stock, it was not possible to design an experiment that studied the effect of changing one geometric parameter in isolation. It is therefore difficult to generalise from such reference points and other techniques need to be explored to define the effects of varying one parameter in isolation. The method described here for providing this additional information was computational fluid dynamics modelling, used initially in a simplified, two-dimensional form to assess the sensitivity of wind loads to changes in height, span and roof pitch. Certain very plausible relationships between geometry and pressures are identifiable from both the full-scale and CFD results, but these trends are not reflected in current wind loading design codes. Some further corroboration is required in order to develop pressure coefficient data suitable for use in design codes which take account of building geometry and so promote efficient design.