In this paper we present a novel approach to the problem of automated design improvement: by treating object-oriented design as a combinatorial optimisation of metrics, we have developed a prototype software engineering tool capable of improving a design with respect to a conflicting set of goals. As the prioritisation of different goals is determined by weights associated with each metric, we also describe here a method of assigning coherent weights to a set of metrics based on object-oriented design heuristics. The combinatorial optimisation approach to automated design improvement is illustrated here by means of a simple case study, which shows the effect of applying our prototype tool to a small inheritance hierarchy. Results indicate that a balance between metrics has been achieved, as several potentially conflicting design goals are accommodated.