The objective of this paper is to present a Generalised Positive Mathematical Programming model suitable for the estimation of variable cost of production associated with different farm activities. This work present, discuss and demonstrates that the Generalised PMP model is a useful theoretical framework for the representation of farm choice, including for the description of costs related to the production function chosen by each entrepreneur. For this characteristics the model can be used for the farms belonging the FADN sample providing a powerful tools for researcher that would like to know variable costs of production for agricultural activities or estimate the impact of agricultural policy and market reform at regional and sectorial level. The main feature of the generalised PMP model is its independence from any “external” information, included the support value of the GME parameters and the abandon of the “tautology” problem always present in the standard PMP models. The paper also present the results of the cost estimation process and validate it comparing the observed variable cost with the estimated variable cost related to a sample of 738 farms belonging the FADN data base of three Italian regions.