Farmers in most OECD countries are engaged in different activities which go beyond agriculture. When assessing farm performance, it is appropriate to model these heterogeneous farm outputs separately. In this study, we use the distance function approach, which allows the consideration of technology with multiple outputs and multiple inputs. We compare estimates from single-output technology with estimates from multiple- output technology. Our empirical analysis is based on an unbalanced panel of dairy farms in the plain region of Switzerland for the period from 2003 to 2009. We choose the parametric estimation method and employ a translog specification of the production technology. The test of output separability favours a model that separately considers three different outputs: agricultural output; para-agricultural output; and direct payments. The separate modelling of direct payments has considerable influence on the estimated technology parameters as well as the technical efficiency scores. The consideration of direct payments as a separate output increases the elasticity of land by a factor greater than two and, accordingly, reduces the distance function elasticities of other inputs. The average technical efficiency estimates do not differ substantially when specifications differ. However, we reveal serious differences in the estimates of technical efficiency for individual farms. The estimated rank correlation coefficients show that the ranking of farms in terms of technical efficiency differs considerably when direct payments are modelled as a separate output.