The common approach used in the thermal design of electrical machines is calibrating thermal models based on the designer's previous experience, or hardware tests on a prototype machine. This allows for various manufacture and assembly nuances to be accounted for in the design process, assuring accurate and computationally efficient predictions of the machine thermal behaviour. The post-manufacture calibration of thermal models from tests on a complete machine has limited use in development of machine topologies, where no previous experience or machine hardware exist. In this context, an experiment informed design technique that makes use of reduced order machine sub-assemblies presents an attractive alternative. In particular, the hardware manufacture cost and time is significantly reduced compared to the prototyping of the complete machine assembly. This allows for numerous hardware samples to be constructed and tested, to inform the machine design process. The use of the machine sub-assembly testing is focused, but not limited to identifying and quantifying various power loss and heat transfer phenomena. This paper reviews the applicability of the sub-assembly testing in a broader context of the machine design. The aim of the research focuses on formulating a basis for sub-assembly based, experiment informed methodology for the thermal design of electrical machines.