This paper addresses issues related to efficiency measurement from an empirical point of view. A stochastic frontier production model using a translog is estimated for a group of Portuguese dairy farms. Farm level survey data for the period 1988 -2005 is used. Stochastic Frontier estimates address behaviour across all periods so the empirical model allows for time varying efficiency as well as technical change. Previous empirical studies use either the value of production or the quantity of milk produced in a single output framework. The value of production approach bundles together decision regarding the quantity and quality components as well as CAP subsidies. Using both approaches for the same data set produce substantial differences in efficiency measurement. The analysis shows that while farmers are quite efficient maximizing quantity produced they are much less efficient when allowing for quality. Sensitivity of estimates to the heterogeneity of the panel data sample as well as to the specification of the dependent variable is discussed. Average efficiency is 84% which indicates a close proximity to the production frontier for 71% of dairy farms but one can not reject the hypothesis that efficiency is decreasing over time and the rate of technical change is negative and close to 2%.