Project-based courses are crucial to gain practically relevant knowledge in modelling and programming education. However, they fall into the "ill-defined" domain: there are many possible solutions; the quality of a deliverable is subjective and not formally assessable; reaching the goals means designing new artefacts and analysing new information; and the problem cannot always be divided into independent tasks. In this paper, we refine the existing two-dimensional (verifiability and solution space) classification of ill-defined classes of problems, contemplate methods and approaches for assessment of projects, and apply the model to analyse two study units of two different computer science programmes.