We discuss individual learning by interactive benchmarking using stochastic frontier models. The interactions allow the user to tailor the performance evaluation to preferences and explore alternative improvement strategies by selecting and searching the different frontiers using directional distance functions. The frontier is given by an explicit quantile, e.g. “the best 90 %”. Using the explanatory model of the inefficiency, the user can adjust the frontiers by submitting state variables that influence the inefficiency. An efficiency study of Danish dairy farms is implemented in the suggested benchmarking tool. The study investigates how different characteristics on dairy farms influences the technical efficiency.