A new approach in hierarchical optimisation is presented which is capable of optimising both the performance and yield of an analogue design. Performance and yield trade offs are analysed using a combination of multi-objective evolutionary algorithms and Monte Carlo simulations. A behavioural model that combines the performance and variation for a given circuit topology is developed which can be used to optimise the system level structure. The approach enables top-down system optimisation, not only for performance but also for yield. The model has been developed in Verilog-A and tested extensively with practical designs using the Spectre simulator. A performance and variation model of a 5 stage voltage controlled ring oscillator has been developed and a PLL design is used to demonstrate hierarchical optimisation at the system level. The results have been verified with transistor level simulations and suggest that an accurate performance and yield prediction can be achieved with the proposed algorithm.