Recent low estimates of total factor productivity change for wool producers in the Australian sheep industry indicate that they are struggling to improve their performance. This evidence is at odds with the views of many technical observers of industry performance, prompting us to re-estimate total factor productivity change for farmers in a benchmarking group in south-west Victoria who had been the subject of such a negative finding. An important transformation in sheep production in Australia in recent decades has been a change in enterprise mix, notably a greater presence of prime lamb production. This change complicates the process of computing efficiency and productivity change. We demonstrate that a multi-input multi-output approach, based on the use of a stochastic output distance function and assuming non-neutral technical change, helps to avoid errors of estimation of productivity change. Following this approach, we find substantial technical progress and productivity gains on farms over the period, 1995/1996 to 2003/2004, for merino wool specialists. Growth in total factor productivity on properties operating wool and prime lamb enterprises is less impressive, indicating an initial decline in total factor productivity in 1996/1997 as farmers adopted a more diversified enterprise mix. Following this decline, however, farms achieved a modest increase in total factor productivity.