To minimize the experimental effort of assessing biaxial incidence angle characteristics of solar collectors with complex geometries, current testing standards (EN 12975-2, ASHRAE 93, ISO 9806:2013) recommend the factorization method proposed by McIntire (1982). Due to a lack of consensus, researchers in the low and high temperature fields of solar thermal have applied this model in very different ways, either referring to different radiation references (beam radiation on tilted surface vs. direct normal irradiance) or angle projections (theta(T) - theta(L)-domain vs. theta(T) - theta(i)-domain). In this study, the influence of each approach on the factorization error is estimated. Four different collector geometries were considered: compound parabolic collector, maximum reflector collector, linear Fresnel reflector, and fixed-mirror collector (Concentrating Collector with Stationary Reflector, CCStaR V2). The 3D incidence angle modifier surface of each collector was calculated with a sophisticated in-house ray tracing tool and compared with the incidence angle modifier surface constructed by various factorization models. The error was defined as difference between direct normal irradiance weighted annual integration of ray tracing and factorized incidence angle modifier, given the collector's location and orientation. This allows the deviations to be represented more intuitively by plotting the relative error over latitude. Factorization in the (theta(T), theta(i))-domain was shown to constantly yield accurate results even in the case of static collectors. (C) 2015 Elsevier Ltd. All rights reserved.