Four dimensional (4D, or 3D plus time) X-ray tomographic imaging of phase changes in materials is quickly becoming an accepted tool for quantifying the development of microstructures to both inform and validate models. However, most of the systems studied have been relatively simple binary compositions with only two phases. In this study we present a quantitative dataset of the phase evolution in a complex three-phase material, ice cream. The microstructure of ice cream is an important parameter in terms of sensorial perception, and therefore quantification and modelling of the evolution of the microstructure with time and temperature is key to understanding its fabrication and storage. The microstructure consists of three phases, air cells, ice crystals, and unfrozen matrix. We perform in situ synchrotron X-ray imaging of ice cream samples using in-line phase contrast tomography, housed within a purpose built cold-stage (-40 to +20oC) with finely controlled variation in specimen temperature. The size and distribution of ice crystals and air cells during programmed temperature cycling are determined using 3D quantification. The microstructural evolution of three-phase materials has many other important applications ranging from biological to structural and functional material, hence this dataset can act as a validation case for numerical investigations on faceted and non-faceted crystal growth in a range of materials.