This paper examines the design of a composite helicopter rotor blade to meet given cross-sectional properties. As with many real-world problems, the choice of objective and design variables can lead to a problem with a non-linear and/or non-convex objective function, which would require the use of stochastic optimisation methods to find an optimum. Since the objective function is evaluated from the results of a finite element analysis of the cross section, the computational expense of using stochastic methods would be prohibitive. It is shown that by choosing appropriate simplified design variables, the problem becomes convex with respect to those design variables. This allows deterministic optimisation methods to be used, which is considerably more computationally efficient than stochastic methods. It is also shown that the design variables can be chosen such that the response of each individual cross-sectional property can be closely modelled by a linear approximation, even though the response of a single objective function to many design parameters is non-linear. The design problem may therefore be reformulated into a number of simultaneous linear equations that are easily solved by matrix methods, thus allowing an optimum to be located with the minimum number of computationally expensive finite element analyses.