This document is Deliverable 7.2 of the Co-UDlabs project, funded under the European Union’s Horizon 2020 research and innovation programme and under Grant Agreement No 101008626. This report describes the historical use of condition grading of pipes and ancillary assets in sewer networks to inform intervention decisions and investment planning. The early condition assessment approaches, used defect data obtained from CCTV inspection to assign a condition grade to an individual pipe. Condition grades were normally integer values. The defect data was obtained by manual inspection of the CCTV images, and the use of a standard defect coding system. The earliest asset management approaches used this asset condition grading to make intervention and investment decision. These condition grading approaches developed more complexity with time, by including information on the consequences of any asset failure and the cost of any intervention. Multi-criteria based schemes were then introduced to explicitly account for socio-economic and environmental factors as well as the physical condition of the assets. The relative weighting of these factors proved to be problematic and different approaches were developed in a number of studies and were applied in different countries. There was little consensus as to value of the relative weighting factors. Two types of uncertainty have been identified, the first associated with the uncertainty associated with mapping uncertain defect data to condition class and the second in terms of the uncertainty in definition and values of the different criteria weighting on the final ranking for intervention and investment. It is clear that current and even emerging inspection technologies will not be able to inspect all assets in the short to medium term. Deterioration models are therefore required to understand the physical condition of individual assets due to this lack of high quality defect data. Three different approaches are currently used, all rely on calibration/training/validation form limited collected condition data. All approaches provide a similar level of predictive performance at a network level, but Machine Learning (ML) based approaches provide more accurate asset condition predictions at an individual asset level. If there was better understanding of the processes that govern the mechanisms that create individual types of defect then the ML based approaches could be applied at a defect level on an individual asset rather than on an aggregated condition grade. This combined with existing hydraulic network modelling approaches would allow the impact of the size and likelihood of failure associated with individual assets to be simulated. This combined with higher resolution social and environmental data would allow better informed multi-criteria decision approaches to be developed and so better identify the individual assets in need of intervention. By working at an individual asset level the assignment of weighting factors to the physical, socio-economic and environmental factors should become more transparent and the only uncertainty that remains is the sensitivity of the final intervention ranking to these weighting values.