To investigate the potential use of a thermochemical software package (FactSage 6.2), in the design of alkaline earth boroaluminosilicate glass ceramics, experimental and modelled results on four glass ceramics were compared. Initially large discrepancies were found. These are described and related to deficiencies in the modelling of borate crystallisation and phase separation from a liquid oxide solution. Furthermore, a disparity in reproducing phase separations reported in the literature, through modelling, was also found. By accounting for potential error sources in the modelling, through semi-empirical optimisation of boron and alkali/alkali earth oxide activities in the liquid oxide solution, significantly improved fits between modelled and experimental results were obtained. Based on these results, it is suggested that more precise descriptions of higher order interactions need to be addressed, to account for the topology of multicomponent melts, before accurate prediction of phase relations within boron-containing glass ceramics can be obtained.