Abstract Thermal cladding systems have developed and modernised since the first systems were implemented, and predictions of single figure sound insulation improvement, ΔRW, based on the natural frequency, f0, of the spring-mass covering may no longer be reliable. To identify aspects of the compound acoustic behaviour due to multiple power flow paths of the thermal insulating system, a statistical energy analysis (SEA) based prediction model was developed. A simplified calculation of sound insulation improvement, ΔR, is described, allowing the high frequency (f > f0) behaviour of thermal cladding systems to be predicted. A parametric study in which the impact of different construction materials in the model is discussed; the damping constants, elastic properties of the interlayer and fixings, number of fixings, thickness and material properties (including bending stiffness) of the weatherproof outer layer and the heavyweight wall are assessed. While agreement within 4.0 dB (mean absolute differences) between calculated and measured results for thick render (≥8.0 mm) and curtain wall systems can be obtained at high frequencies (f > f0) using the simplified methodology, this approach was not successful at predicting single figure values. This is because single figure values are weighted towards the low frequencies. Correlation of calculated f0 with measured ΔRW is slightly improved (r.m.s. differences of 2.62 compared with 3.21 using the f0 calculation methodology in EN ISO12354 Annex D) when a modified method to calculate the combined stiffness is used. To improve predictions further, a methodology must be developed to obtain the transfer function, Ytr, used to calculate non-resonant coupling loss factor due to the spring-mass resonance of thermal cladding on the heavyweight wall. The mobility of the connections, Yc, should also be accurately characterised to ensure accurate predictions at high frequencies.