2026 Volume 7 Issue 3 Pages 339-345
Operational data collected from construction machinery at work sites are observational data affected by changing conditions, such as haul roads, road surface conditions, work contents, and operator assignment. Therefore, the effect of digital support tools cannot be appropriately evaluated by simple before-and-after comparisons. This study examines an evaluation procedure based on the framework of statistical causal inference, using a dump truck operation support application as a case study. The procedure e valuates the influence of a new application function on a cycle-level fuel efficiency indicator. In addition, sensitivity analysis for unobserved confounding is applied to assess the robustness of the observed effect. By presenting both the observed effect size and the robustness of the corresponding judgment, this study discusses the applicability of the proposed approach to explainable decision support based on practical operational data.