A method of operating a communications network is disclosed. Modern communications networks produce vast amounts of network operational data which have the potential to provide a useful summary of the operational state of the network. Whilst processes such as clustering are known for arranging the vast amount of data into groups, the clusters themselves do not provide data which might be easily interpreted by network elements or administrators. Network operational data often comprises a plurality of data items, each of which gives a value for each of a set of attributes. By processing a cluster to identify attributes in the cluster whose values vary less in the cluster then they vary outside of the cluster, and then generating a cluster description which is based on a measure of the central tendency of the values of those attribute in the cluster, an easily interpretable general description of the data items in the cluster is provided. The easily interpretable general description of the cluster can then be used to relatively identify data items similar to those present in the cluster (e.g. from a larger database of data items), and elements in the network can then act autonomously on the basis of the cluster description to control the operation of the communications network.