As for the asset management of existing infrastructures, making a decision objectively for repair timings is an important issue, and the statistic deterioration prediction based-upon visual inspection data evaluated by multi-staged ratings will be one of the efficient way. However, most of infrastructure managers do not possess enough inspection data to carry out the statistic deterioration prediction with high accuracy. This study addresses a Bayesian updating methodology which can provides the prior estimates of the deterioration process by the expert’s subjective and empirical judgments at the early stages, and revises them sequentially based on the additional data obtained through inspections. The proposed method is applied to the actual visual inspection data for painting deterioration of steel girders to verify the effects of the prior subjective information to Bayesian updating results of deterioration prediction.