— This paper presents a complete solution to the problem of how to parametrise cluster-based stochastic MIMO channel models from measurement data, with minimum user intervention. The method comprises the following steps: (i) identify clusters in measurement data, (ii) identify the optimum number of clusters, (iii) track clusters over consecutive time snapshots, (iv) estimate cluster parameters. These parameters are given as estimated probability density functions of the cluster power, cluster positions, delay and angular spreads of clusters and the number of paths within a cluster. Applied to high-resolution indoor MIMO measurement data at 5.2 GHz and at 2.55 GHz, the method yields coherent results of the obtained cluster parameters.