Designing constructions to provide sound insulation typically involves the use of measured data along with statistical and/or analytical prediction models, blended with a combination of empiricism, experience and pragmatism. In many situations, the accuracy of input data that is determined from laboratory or field measurements is often difficult to assess without recourse to other prediction models. This is evident in the prediction of flanking transmission where input data is often needed from measurements of structure-borne sound transmission across junctions of walls and floors. In fact, the complexity of many junctions means that we sometimes need to consider parts of a building as a black box where we only consider the input and output variables; this provides a practical solution for vibration transmission across the foundations of masonry cavity walls or complex timber-frame constructions. Historically, the development of prediction models has tended to focus only on steady-state sources because of the procedures that are commonly used to measure airborne sound insulation, and impact sound insulation with the tapping machine. However, for transient or time-varying sources, regulations and Standards require measurement of Fast time-weighted maximum sound pressure levels, for example with impact sound insulation measured using the rubber ball as a heavy impact source, or noise generated by building machinery. To address this need, models have been developed and validated based on Transient SEA (TSEA) along with an empirical approach that could be potentially be incorporated in EN ISO 12354. In the laboratory, we rely on measurements made with artificial ‘heavy’ and ‘intense’ rain from a single raindrop diameter to assess rain noise on roofs and roof glazing. To interpret these measurements for design purposes, models have been developed to translate the force applied by individual drops of water to natural rainfall which has a range of drop diameters. Challenges still remain in designing for low-frequency performance. Despite the fact that measurement methods for the laboratory and the field are available to improve the repeatability, reproducibility and relevance of low-frequency sound insulation (below 100Hz), they are not widely used. In addition, models such as EN ISO 12354 do not predict the low-frequency sound insulation for a single receiving room in a building where the mode count in one-third octave bands is low for the rooms, walls and floors.