Random sampling of capacities is an important component of various approaches to learning and optimisation, especially in problems involving complex constraints and objectives. Being able to generate unbiased random capacities is also crucial when it comes to validation and evaluating comparative performance of newly proposed capacity identification methods on randomly generated datasets. In this chapter we summarise the main challenges in random sampling, as well as a number of useful approaches to random capacity generation. Some of these methods apply to general capacities, while others are specialised towards sparse capacities and specific types.