Advancing readiness and nonlinear pathways to infrastructure project success through ‘ glass box ’ machine learning models

We examine how operations readiness affects infrastructure project success using an interpretable machine learning model built on survey data from 633 operations and project professionals. It assesses how ‘Facility’, ‘People’, ‘Technology’ and ‘Organisation’ readiness shape success, and was benchmarked against ten other models. A combined model fitted the training data best, but the interpretable model predicted new cases most accurately, with simple linear regression close behind; its main value lies in explanation rather than prediction. ‘Organisation’ readiness was the strongest predictor. ‘People’ readiness delivered early gains that then levelled off. Both patterns held under repeated resampling. ‘Technology’ readiness showed a threshold effect, though whether its benefits level off thereafter is uncertain. The ‘Facility’ readiness threshold location and the absence of diminishing returns for ‘Organisation’ readiness are also tentative. The strongest interaction linked ‘Technology’ and ‘Organisation’ readiness, giving managers sample-specific guidance on prioritising readiness investments.