Counterfactual Modelling for Evaluating Pipeline Replacement Strategies

Abstract. Water utilities companies increasingly rely on predictive models to prioritise pipeline replacement, yet they rarely evaluate whether these decisions actually prevent failures. This study applies a counterfactual modelling approach to assess the effectiveness of Brabant Water’s replacement strategy in the Netherlands. Using rare-events logistic regression on pipeline and environmental characteristics, failure probabilities were estimated for pipelines replaced between 2020 and 2023 and simulated a no-replacement scenario. Comparing simulated counterfactual failures with observed post-replacement failures shows that executed replacements reduced failure rates by more than 80%. Spatial hotspot analysis further reveals that most simulated high-risk areas did not experience failures after replacement, indicating successful risk mitigation. The study demonstrates that counterfactual reasoning provides a powerful framework for evaluating infrastructure policies, provided that predictive models are well calibrated.