Recent studies have pointed out the general tendency of scale models to overestimate the daylighting performance of buildings, usually expressed through work plane illuminance and daylight factor distribution profiles. An analysis of the corresponding sources of error has allowed to identify the main parameters responsible for the overestimation - such as indoor surfaces reflectance, glazing transmittance and photometers features. It was shown that a careful mock-up of the real building characteristics can reduce the divergence of the scale models daylighting performance down to 30 %, even for locations situated away from the window side. An appropriate tuning of the numerical parameters involved in daylighting computer simulation models lead to comparable accuracies, as shown by different authors. Daylighting computer simulations of a real building (a 1:1 scale daylighting test module), together with a virtual model of the corresponding 1: 10 scale model placed in a scanning sky simulator, were used to carry out an in-depth analysis of the sources of error of both physical and virtual modeling techniques. Through a computer sensitivity analysis of the most significant parameters influencing the accuracy of the physical model, design rules and error calculation methods for scale models are expected to be drawn.