The primary objective of the project was the development of criteria to define the end-of–life condition of pavements. These criteria could then be used in pavement performance modelling to obtain a more robust measure of remaining life. Another objective was the generation of a new model for maintenance costs. This could then be combined with the existing models for roughness and rutting to define a distress level at which rehabilitation should occur. None of the maintenance cost models developed were particularly successful in producing a reliable prediction of maintenance costs based on the pavement characteristics available from RAMM. Therefore, a logit model was developed to predict rehabilitation decisions. The major factors in the rehabilitation model were maintenance costs, traffic levels and roughness. The rehabilitation decision model derived for this study predicted rehabilitation decisions well. Approximately 72% of pavements that had been rehabilitated were predicted as requiring rehabilitation. When tested on the Nelson network data, which was not used for calibration of the model, a similar performance was obtained indicating the models developed were relatively successful.