In this report, a brief overview is given on how to model deterioration for the purpose of maintenance optimization. Without being complete, we presented reliability-based models and condition-based models developed for pipes, bridges and roads. In the field of reliability-based models, survival functions are the favourite approach when trying to balance model output and input: results are meaningful and the possibility of aggregating asset in homogeneous populations reduces the data requirements. The survival functions have been tested for pipes and bridges; the application in the water sector has been quite successful. However, when asset can be inspected, the condition – based models are a favourite solution since they are based on direct inspection of defects: the combination of historical data with Markov-chain methodology seems to be a feasible approach if enough data is available. The condition and reliability deterioration may have little or no correlation even if they refer to the same element. In fact, an element may fail suddenly even if its visual conditions before failure is perfect, or may be extremely safe even if its visual conditions is extremely poor. The system serviceability can only be captured by identifying a link between system condition and performance and then evaluating system reliability based on acceptable levels of performance.The report discusses the advantages and limitations of different models and draw conclusions on how to proceed within the SIP project.