Predictive maintenance is a powerful tool to prevent unplanned failures and ensure that operational plans and budgets are met. Any prediction carries uncertainty, and traditionally we have described this uncertainty using measures of uncertainty, such as p-values and confidence intervals. However, these can be easily misinterpreted. We discuss the interpretation of a common modelling approach (proportional hazards) from Bayesian and frequentist perspectives. Many industries are keen to adopt Bayesian statistics for predictive maintenance but have historically been reluctant to do so without convenient software packages. Recent developments in open source code and computational capacity have changed this situation. As a result, use cases exploring the potential value of Bayesian approaches on in-service reliability case studies are needed. We demonstrate the use of the Bayesian analysis of a Weibull proportional hazards model using blockage data for 44,800 vitrified clay wastewater pipes provided by a water utility. The data includes a variety of numerical and categorical covariates with varying sample sizes. Due to the nature of this data set, the estimates for hazard ratios for each covariate are quite similar. Rather than a p-value as the main frequentist measure of uncertainty, under the Bayesian approach, we report our results using probability values that are familiar to us. The Bayesian approach also allows us to predict a mean time to failure for a