Property portfolios are traditionally constructed by diversifying across geographical areas, property types or a combination of both. In the UK it is normal practice to use regions rather than towns or local markets areas as the geographical divisions. In this paper cluster analysis is used to construct homogeneous groups from 157 UK local markets using commercial property returns. The results show strong property type dimensions and only very broad geographical dimensions in the clusters. These clusters are found, generally, to have temporal stability with changes in cluster membership explained by the changing economic geography of the UK. The cluster-derived groupings are used to derive efficient investment frontiers and are compared to frontiers based on conventional heuristic groupings. It is shown that strategies based on parsimonious cluster-based groupings, appropriate for smaller investors, generate results that are comparable to those of conventional groupings and capture the main drivers of property performance.