We report on a theoretical framework for magnetic hyperthermia where the amount of heat generated by nanoparticles can be understood when both the physical and hydrodynamic size distributions are known accurately. The model is validated by studying the magnetic, colloidal and heating properties of magnetite/maghemite nanoparticles of different sizes dispersed in solvents of varying viscosity. We show that heating arising due to susceptibility losses can be neglected with hysteresis loss being the dominant mechanism. We show that it is crucial to measure the specific absorption rate of samples only when embedded in a solid matrix to avoid heating by stirring. However the data shows that distributions of both size and anisotropy must be included in theoretical models.