Metal accounting is one of the main tools for financial and technical management of metal production industry. It is based on measurements and has to manage the uncertainty inherent to the measurement process. The uncertainty in the metal accounting generates financial risk. The accuracy of the metal accounting results is directly linked to the accuracy of the material balance and then to the accuracy of the mass and content measurements. Estimate the overall measurement error, through its probability distribution or its first and second moments (mean and variance), can contribute to the enterprise decision making. The overall measurement error can be calculated and analysed by establishing the uncertainty budget. If this approach has been mainly introduced to calculate the analytical error (cf. ISO GUM), it has to take into account the sampling procedure. Even though it is not explicitly named “uncertainty budget”, the same approach is proposed in the Pierre Gy’s Theory of Sampling (TOS), where the various components of the overall error are well identified and described with their properties and their relative weights. The present paper proposes a methodology to build such uncertainty budgets in the frame of the implementation of a metal accounting system. It can be applied to an existing measurement system, analysing the results in order to find some ways for improving the measurement accuracy. In addition, it can be used to define a new measurement procedure with an objective of accuracy. Various real examples illustrate both applications.