When designing a solar photovoltaic (PV) system, the main aim of the designer is to optimise the system through the correct choice in sizing of system components (i.e.: size of PV array versus number of batteries:for storage), whilst at the same time keeping the overall systems costs as low as possible. The choice of data provider and method for measuring the solar irradiance data in W/m(2) for a specific geographical location is therefore a critical determinant to ensure maintenance of the desired level of accuracy for the system design. The use of a data set which either over predicts or under predicts the amount of available solar irradiation at a specific location will therefore affect the electrical performance of the system, as the real world conditions may differ considerably to the data set used in the modelling of the system design. This paper specifically deals with the modelling of an off-grid photovoltaic power supply system using three different sources of solar irradiance data for two specific geographical locations within :South Africa. The principal aim of the paper is to show that satellite data over-estimates solar irradiance and hence underestimates the initial cost of a solar PV system needed to meet the load in question. (C) 2012 Elsevier Ltd. All rights reserved.