A split-panel latent class stochastic frontier model is applied to account for technological heterogeneity among Swiss dairy farms and to assess the potential performance improvements through technology choice and change over 11 years. Three technology classes with substantially different productivity levels are identified considering the unobserved and observed farm characteristics. Technologies seem on average well adapted to local natural production conditions with low potential for efficiency and productivity increases. Few farms changed technology over time and either an intensification or extensification strategy was observed. Policymakers must be aware of the interlinkages between technology choices and the economic situation of farms.