As data from monitored structures become more available, the demand for its efficient use in structural operation and management grows. This can be achieved by using structural response measurements to assess the usefulness of models describing deterioration processes and the mechanical behaviour of structures. This work aims to frame Structural Health Monitoring as a Bayesian model updating problem, where the quantities of inferential interest characterise the deterioration process and/or structural state. Using posterior estimates of these quantities, a decision-theoretic definition is proposed to assess the models based on (a) their ability to explain the data and (b) their performance in decision support tasks. The framework is demonstrated on strain response data from a test specimen subjected to three-point bending and accelerated corrosion, leading to thickness loss. Results indicate that a priori domain knowledge of the deterioration form is critical.