One key aspect of sewer inspection programs is the prediction of sewer condition. Despite the development of deterioration models, the influence of available data on models' predictive power has not been studied in depth. In this article, numerical experiments on a semi-virtual asset stock have been conducted to answer two main questions: how to establish a list of the most informative factors and whether it is better to have data imprecision instead of data incompleteness in a utility database. Two approaches for establishing a list of the most informative factors are compared. The results show a statistical analysis (a priori analysis) can predict the impact of available data on inspection program efficiency (a posteriori analysis). This can be used to plan data acquisition programs. Finally, we show that using the notion of "district" (data imprecision) can provide efficient results when the most informative factor "age" is not available (data incompleteness).