Invariants play a crucial role in system development. This contribution focuses on invariants in systems with so-called occurrence uncertainty, where we are interested in deciding whether a certain population (a set of instances of a class model) of the system satisfies an invariant or not, but we are unsure about the actual occurrence of the elements of that population, and also about the degree of satisfaction that is actually required for the invariant to be fulfilled. Invariants are soft in the sense that they are required to hold only for a particular, and a priori uncertain, percentage of the population. The contribution proposes a systematic approach to occurrence uncertainty and a prototypical implementation for models with uncertainty and soft invariants allowing to build system states and to make experiments with them.