A standard TPS measurement gives the thermal conductivity and thermal diffusivity of an isotropic material which in turn gives the heat capacity. The thermal properties of an anisotropic material can be measured if the heat capacity is known. A method for heat capacity measurement exists, where the TPS sensor is attached to a sample container which is surrounded by insulation. However, it's based on an assumption of negligible heat losses which leads to uncertainties in the results. From that position, this work aims to model the heat losses from the specific heat measurements with TPS. A new set-up is introduced, where the sensor with the container hangs freely in a steel tube to get more predictable heat losses. The results show that the measurements can be modelled as a network of lumps connected by conductances approximated as constant. Thereby, the conductances out from the system can be solved from a reference measurement and used as input for a model of a measurement with a sample. The model seems to underestimate the heat capacity, which might be a consequence of temperature dependent effects on the conductances from convection. The tube in the set-up could be evacuated to minimize those losses.