Ultrasonic Injection moulding has emerged as an alternative production route for miniature and microscale polymeric components, where it offers some significant benefits over conventional microinjection processes, including reduced residence time at high temperature, lower required injection pressures, lower energy usage and easier cleaning. Heating of the polymer prior to injection is achieved entirely using applied ultrasonic energy using a 30 KHz source which is bought into contact with a predefined number of pellets in the sonication chamber. Heating of these pellets is generated from two distinct mechanisms, interfacial friction heating (caused by frictional heating at the contact points between the pellets) and viscoelastic heating (caused by heat generated by internal friction in the material, due to damping of the oscillations by the polymer itself). One challenge associated with these mechanisms is to ensure a repeatable moulding environment in the machine, because heating can be strongly influenced by the initial frictional heating, which itself is dependent on the size, shape and number of pellets and the resulting contact surface areas during the early phases of the heating process. The pellets are dropped into this chamber in an essentially random manner, so small changes in this initial configuration can affect heat generation and cause significant differences between consecutive moulding cycles. In this work we investigate the significance of these early events by controlling the initial contact areas using a range of defined particulate geometries including a stack of thin discs and uniform spheres. We quantify the uniformity of the temperature distribution for each shot and from shot to shot using a novel injection mould tool which allows direct viewing of the mould cavity using high speed conventional and infrared imaging. The uniformity of the resulting physical properties has been examined using small scale mechanical testing apparatus.